Meaning in Mathematics - John Polkinghorne
Meaning in Mathematics - John Polkinghorne
Meaning in Mathematics - John Polkinghorne
Edited by
John Polkinghorne
1
3
Great Clarendon Street, Oxford OX2 6DP
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide in
Oxford New York
Auckland Cape Town Dar es Salaam Hong Kong Karachi
Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto
With offices in
Argentina Austria Brazil Chile Czech Republic France Greece
Guatemala Hungary Italy Japan Poland Portugal Singapore
South Korea Switzerland Thailand Turkey Ukraine Vietnam
Oxford is a registered trade mark of Oxford University Press
in the UK and in certain other countries
Published in the United States
by Oxford University Press Inc., New York
© Oxford University Press 2011
The moral rights of the authors have been asserted
Database right Oxford University Press (maker)
First published 2011
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
without the prior permission in writing of Oxford University Press,
or as expressly permitted by law, or under terms agreed with the appropriate
reprographics rights organization. Enquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the address above
You must not circulate this book in any other binding or cover
and you must impose the same condition on any acquirer
British Library Cataloguing in Publication Data
Data available
Library of Congress Cataloging in Publication Data
Library of Congress Control Number: 2011920646
Typeset by SPI Publisher Services, Pondicherry, India
Printed in Great Britain
on acid-free paper by
Clays Ltd, St Ives plc
ISBN 978–0–19–960505–7
1 3 5 7 9 10 8 6 4 2
In grateful memory of Peter Lipton, scholar and friend.
This page intentionally left blank
Contents
List of contributors ix
Introduction 1
John Polkinghorne
3 Mathematical reality 27
John Polkinghorne
Comment 35
Mary Leng
Reply 39
John Polkinghorne
References 147
Index 153
List of contributors
The title of this chapter is a famous question. Indeed, perhaps it is a little too
famous: it has been asked over and over again, and it is not clear what would
constitute a satisfactory answer. However, I was asked to address it during the
discussions that led to this volume, and since most of the participants in those
discussions were not research mathematicians, I was in particular asked to give
a mathematician’s perspective on it.
One reason for the appeal of the question seems to be that people can use
it to support their philosophical views. If mathematics is discovered, then it
would appear that there is something out there that mathematicians are discov-
ering, which in turn would appear to lend support to a Platonist conception of
mathematics, whereas if it is invented, then that might seem to be an argument
in favour of a non-realist view of mathematical objects and mathematical truth.
But before a conclusion like that can be drawn, the argument needs to be
fleshed out in detail. First, one must be very clear what it means to say that
some piece of mathematics has been discovered, and then one must explain,
using that meaning, why a Platonist conclusion follows. I do not myself believe
that this programme can be carried out, but one can at least make a start on
it by trying to explain the incontestable fact that almost all mathematicians
who successfully prove theorems feel as though they are making discoveries.
It is possible to think about this question in a non-philosophical way, which
is what I shall try to do. For instance, I shall consider whether there is an
identifiable distinction between parts of mathematics that feel like discoveries
and parts that feel like inventions. This is partly a psychological question and
partly a question about whether there are objective properties of mathematical
statements that explain how they are perceived. The argument in favour of
Platonism only needs some of mathematics to be discovered: if it turns out that
there are two broad kinds of mathematics, then perhaps one can understand
the distinction and formulate more precisely what mathematical discovery (as
opposed to the mere producing of mathematics) is.
4 M E A N I N G I N M AT H E M AT I C S
As the etymology of the word ‘discover’ suggests, we normally talk of
discovery when we find something that was, unbeknownst to us, already there.
For example, Columbus is said to have discovered America (even if one can
question that statement for other reasons), and Tutankhamun’s tomb was dis-
covered by Howard Carter in 1922. We say this even when we cannot directly
observe what has been discovered: for instance, J. J. Thompson is famous as
the discoverer of the electron. Of greater relevance to mathematics is the dis-
covery of facts: we discover that something is the case. For example, it would
make perfectly good sense to say that Bernstein and Woodward discovered (or
contributed to the discovery) that Nixon was linked to the Watergate burglary.
In all these cases, we have some phenomenon, or fact, that is brought
to our attention by the discovery. So one might ask whether this transition
from unknown to known could serve as a definition of discovery. But a few
examples show that there is a little more to it than that. For instance, an
amusing fact, known to people who like doing cryptic crosswords, is that
the words ‘carthorse’ and ‘orchestra’ are anagrams. I presume that somebody
somewhere was the first person to notice this fact, but I am inclined to call it an
observation (hence my use of the word ‘notice’) rather than a discovery. Why
is this? Perhaps it is because the words ‘carthorse’ and ‘orchestra’ were there
under our noses all the time and what has been spotted is a simple relationship
between them. But why could we not say that the relationship is discovered
even if the words were familiar? Another possible explanation is that once the
relationship is pointed out, one can very easily verify that it holds: you don’t
have to travel to America or Egypt, or do a delicate scientific experiment, or
get access to secret documents.
As far as evidence for Platonism is concerned, the distinction between
discovery and observation is not especially important: if you notice something,
then that something must have been there for you to notice, just as if you
discover it then it must have been there for you to discover. So let us think of
observation as a mild kind of discovery rather than as a fundamentally different
phenomenon.
How about invention? What kinds of things do we invent? Machines are
an obvious example: we talk of the invention of the steam engine, or the
aeroplane, or the mobile phone. We also invent games: for instance, the British
invented cricket—and more to the point, that is an appropriate way of saying
what happened. Art supplies us with a more interesting example. One would
never talk of a single work of art being invented, but it does seem to be
possible to invent a style or a technique. For example, Picasso did not invent
Les Desmoiselles d’Avignon, but he and Braque are credited with inventing
cubism.
A common theme that emerges from these examples is that what we
invent tends not to be individual objects: rather, we invent general methods
for producing objects. When we talk of the invention of the steam engine, we
are not talking about one particular instance of steam-enginehood, but rather of
the idea—that a clever arrangement of steam, pistons, etc. can be used to drive
IS M AT H E M AT I C S D I S C O V E R E D O R I N V E N T E D ? 5
and many other facts that can be used to build up the theory of complex
numbers. A second approach, which was introduced much later in order to
demonstrate that the complex number system was consistent if the real number
system was, is to define a complex number to be an ordered pair (a, b) of real
numbers, and to stipulate that addition and multiplication of these ordered pairs
are given by the following rules:
(a, b) + (c, d) = (a + c, b + d)
(a, b)(c, d) = (ac − bd, ad + bc)
This second method is often used in university courses that build up the number
systems rigorously. One proves that these ordered pairs form a field under the
two given operations, and finally one says, ‘From now on I shall write a + bi
instead of (a, b).’
10 M E A N I N G I N M AT H E M AT I C S
Another reason for our ambivalence about the complex numbers is that
they feel less real than real numbers. (Of course, the names given to these
numbers reflect this rather unsubtly.) We can directly relate the real numbers
to quantities such as time, mass, length, temperature, and so on (though for this
we never need the infinite precision of the real number system), so it feels as
though they have an independent existence that we observe. But we do not run
into the complex numbers in that way. Rather, we play what feels like a sort of
game—imagine what would happen if −1 did have a square root.
But why in that case do we not feel happy just to say that the complex
numbers were invented? The reason is that the game is much more interesting
than we had any right to expect, and it has had a huge influence even on
those parts of mathematics that are about real numbers or even integers. It
is as though after our one small act of inventing i, the game took over and
we lost control of the consequences. (Another example of this phenomenon is
Conway’s famous game of Life. He devised a few simple rules, by a process
that one would surely want to regard as closer to invention than discovery, but
once he had done so he found that he had created a world full of unexpected
phenomena that he had not put there, so to speak. Indeed, most of them were
discovered—to use the obvious word—by other people.)
Why is ‘discovery of non-Euclidean geometry’ more popular than ‘inven-
tion of non-Euclidean geometry’? This is an interesting case, because there
are two approaches to the subject, one axiomatic and one concrete. One could
talk about non-Euclidean geometry as the discovery of the remarkable fact
that a different set of axioms, where the parallel postulate is replaced by a
statement that allows a line to have several parallels through any given point,
is consistent. Alternatively, one could think of it as the construction of models
in which those axioms are true. Strictly speaking, one needs the second for the
first, but if one explores in detail the consequences of the axioms and proves
all sorts of interesting theorems without ever reaching a contradiction, that can
be quite impressive evidence for their consistency. It is probably because the
consistency interests us more than the particular choice of model, combined
with the fact that any two models of the hyperbolic plane are isometric, that we
usually call it a discovery. However, Euclidean geometry (wrongly) feels more
‘real’ than hyperbolic geometry, and there is no single model of hyperbolic
geometry that stands out as the most natural one; these two facts may explain
why the word ‘invention’ is sometimes used.
My final example was that of proofs, which I claimed could be discovered
or invented, depending on the nature of the proof. Of course, these are by
no means the only two words or phrases that one might use: some others
are ‘thought of’, ‘found’, ‘came up with’. And often one regards the proof
less as an object than as a process, and focuses on what is proved, as is
shown in sentences such as, ‘After a long struggle, they eventually managed
to prove/establish/show/demonstrate that . . . ’ Proofs illustrate once again the
general point that we use discovery words when the author has less control and
invention words when there are many choices to be made. Where, one might
IS M AT H E M AT I C S D I S C O V E R E D O R I N V E N T E D ? 11
ask, does the choice come from? This is a fascinating question in itself, but let
me point out just one source of choice and arbitrariness: often a proof requires
one to show that a certain mathematical object or structure exists (either as
the main statement or as some intermediate lemma), and often the object or
structure in question is far from unique.
Before drawing any conclusions from these examples, I would like to
discuss briefly another aspect of the question. I have been looking at it mainly
from a linguistic point of view, but, as I mentioned right at the beginning, it
also has a strong psychological component: when one is doing mathematical
research, it sometimes feels more like discovery and sometimes more like
invention. What is the difference between the two experiences?
Since I am more familiar with myself than with anybody else, let me draw
on my own experience. In the mid 1990s I started on a research project that
has occupied me in one way or another ever since. I was thinking about a
theorem that I felt ought to have a simpler proof than the two that were then
known. Eventually I found one (here I am using the word that comes naturally);
unfortunately it was not simpler, but it gave important new information. The
process of finding this proof felt much more like discovery than invention,
because by the time I reached the end, the structure of the argument included
many elements that I had not even begun to envisage when I started working on
it. Moreover, it became clear that there was a large body of closely related facts
that added up to a coherent and yet-to-be-discovered theory. (At this stage,
they were not proved facts, and not always even precisely stated facts. It was
just clear that ‘something was going on’ that needed to be investigated.) I and
several others have been working to develop this theory, and theorems have
been proved that would not even have been stated as conjectures fifteen years
ago.
Why did this feel like discovery rather than invention? Once again it is
connected with control: I was not selecting the facts I happened to like from
a vast range of possibilities. Rather, certain statements stood out as obviously
natural and important. Now that the theory is more developed, it is less clear
which facts are central and which more peripheral, and for that reason the
enterprise feels as though it has an invention component as well.
A few years earlier, I had a different experience: I found a counterexample
to an old conjecture in the theory of Banach spaces. To do this, I constructed
a complicated Banach space. This felt partly like an invention—I did have
arbitrary choices, and many other counterexamples have subsequently been
found—and partly like a discovery—much of what I did was in response to the
requirements of the problem, and felt like the natural thing to do, and a very
similar example was discovered independently by someone else (and even the
later examples use similar techniques). So this is another complicated situation
to analyse, but the reason it is complicated is simply that the question of how
much control I had is a complicated one.
What conclusion should we draw from all these examples and from how
we naturally seem to regard them? First, it is clear that the question with which
12 M E A N I N G I N M AT H E M AT I C S
we began is rather artificial. For a start, the idea that either all of mathematics
is discovered or all of mathematics is invented is ridiculous. But even if we
look at the origins of individual pieces of mathematics, we are not forced to
use the word ‘discover’ or ‘invent’, and we very often don’t.
Nevertheless, there does seem to be a spectrum of possibilities, with some
parts of mathematics feeling more like discoveries and others more like inven-
tions. It is not always easy to say which are which, but there does seem to be
one feature that correlates strongly with whether we prefer to use a discovery-
type word or an invention-type word. That feature is the control that we have
over what is produced. This, as I have argued, even helps to explain why the
doubtful cases are doubtful.
If this is correct (perhaps after some refinement), what philosophical con-
sequences can we draw from it? I suggested at the beginning that the answer
to the question did not have any bearing on questions such as ‘Do numbers
exist?’ or ‘Are mathematical statements true because the objects they mention
really do relate to each other in the ways described?’ My reason for that
suggestion is that pieces of mathematics have objective features that explain
how much control we have over them. For instance, as I mentioned earlier, the
proof of an existential statement may well be far from unique, for the simple
reason that there may be many objects with the required properties. But this
is a straightforward mathematical phenomenon. One could accept my analysis
and believe that the objects in question ‘really exist’, or one could view the
statements that they exist as moves in games played with marks on paper, or
one could regard the objects as convenient fictions. The fact that some parts of
mathematics are unexpected and others not, that some solutions are unique and
others multiple, that some proofs are obvious and others take a huge amount
of work to produce—all these have a bearing on how we describe the process
of mathematical production and all of them are entirely independent of one’s
philosophical position.
Comment on Timothy Gowers’ ‘Is
mathematics discovered or invented?’
Gideon Rosen
This strikes me as exactly right, but it raises a question that Gowers does not
address. Gowers has described the conditions under which mathematicians
are inclined to say that some achievement amounts to a discovery or an
invention, and also the conditions under which an achievement is likely to
feel like a discovery or an invention to those whose achievement it is. But how
seriously should we take these linguistic and psychological observations? As
14 M E A N I N G I N M AT H E M AT I C S
philosophers have often noted, it is one thing to chart the circumstances under
which we are inclined to say this or that, another to identify the conditions
under which it is literally correct to say this or that. So let us grant that
mathematicians agree in their classification of some episode as a matter of
(say) invention. Does that entail or even suggest that this episode was in fact a
matter of invention? Or is this rather a mere manner of speaking that it would
be a mistake to take too seriously?
I believe that this question has different answers in different cases. As
Gowers notes, we speak of the invention/discovery of many kinds of thing:
theories, theorems, proofs and proof techniques, but also mathematical objects
of various sorts (numbers, number systems). We can say that Cantor invented
the theory of transfinite numbers, but we are much less likely to say that Cantor
invented the transfinite numbers themselves. Let’s focus first on the rhetoric
of invention/construction as applied to mathematical objects. Here Gowers
discusses the case of the Monster group, an enormous finite group whose
existence and uniqueness were proved in 1982 and 1990, respectively. The
linguistic evidence suggest that mathematicians are more inclined to speak of
the ‘construction’ of the Monster group than of its ‘discovery’, and Gowers’
account explains this. The proof of the existence of the Monster group is not
unique: many examples may be adduced to establish the existential theorem
that a group with the relevant properties exists, even though (as it happens)
every such example is isomorphic to every other. But is there any reason to
take the imagery of construction seriously in this case? In my view it is a
non-negotiable feature of the literal use of this idiom that if a thing has been
invented or constructed, it did not exist before it was invented and would
not have existed if it had not been invented. By contrast, when a thing is
discovered, it must exist prior to (or at least independently of) the episode
of discovery. But as I think Gowers would agree, it would be quite odd to say
that before 1982, the Monster group did not exist. If this were the right thing
to say, then when Griess first asked himself the question, ‘Does the Monster
exist?’ the answer should have been obvious: ‘Not yet, but maybe someday.’
But in fact no one speaks of mathematical objects in this way. I am therefore
tempted to conclude that even if Gowers is right about the conditions under
which we are inclined to reach for the language of invention or construction
in connection with mathematical objects, it would be a mistake to take this
language literally in this connection.
Things are otherwise when it comes to mathematical theories—especially
large theories like the calculus. If someone had asked in (say) 1650 whether
there existed a powerful battery of algebraic techniques for calculating the area
bounded by a curve and the line tangent to a curve at a point, and a deep theory
justifying these techniques and displaying their connections, the answer might
well have been, ‘Not yet, but maybe someday.’ Moreover, it seems equally
natural to say that if no one had ever managed to write down such a theory, then
the calculus, as we know it, would not exist. Theories of this sort thus appear to
belong to the same ontological category as novels and poems and philosophical
IS M AT H E M AT I C S D I S C O V E R E D O R I N V E N T E D ? C O M M E N T 15
treatises. Such things are abstract artifacts: abstract entities that come into
existence when someone produces a concrete representation of them for the
first time. In these cases, I see no reason not to take the rhetoric of invention
seriously as a sober and literal account of the underlying metaphysics.
Gowers makes no claims of this sort, but I wonder, however, whether he
would agree with my assumption that unless we are prepared to say that the
invented item did not exist prior to its invention, we should regard claims of
invention (construction, creation, etc.) in mathematics as metaphorical. We
might then take Gowers’ careful account of the conditions under which we are
inclined to deploy the metaphor as an account of the sober and metaphysically
neutral truth that underlies it.
This page intentionally left blank
2
Exploring the mathematical library
of Babel
Marcus du Sautoy
Fig. 2.2 Two walls in the Alhambra with the same group of symmetries called 632.
Professor du Sautoy reconciles the realist and the constructivist positions in the
philosophy of mathematics with a simple, but effective distinction. Structures
describable in mathematical language exist independently of our knowledge;
this is the realist part. The mathematician chooses, from among these struc-
tures, those which are to be called mathematical structures. To be describable
in mathematical language is not yet to be a mathematical structure. Professor
du Sautoy adds that aesthetical considerations play a dominant role in deciding
what is worth investigating, i.e., what is to be called mathematics. This is
exactly the position I took in my book, The Applicability of Mathematics
as a Philosophical Problem (1998). The way I put it is that mathematics is
anthropocentric to the extent that anthropocentric criteria (like aesthetical)
govern what is called mathematics.
What now of Hardy’s view that beautiful mathematics is never ‘useful’,
which du Sautoy quotes approvingly? I do not see any reason for Professor
du Sautoy to accept what is a patently false view, based mainly on wishful
thinking. (Hardy didn’t want mathematics to be used for warfare.) Hardy is so
spectacularly wrong that, on the contrary, many scientists are convinced that
the more beautiful mathematics is, the more applications it has. Hardy wrote,
‘No one has yet discovered any warlike purpose to be served by the theory
of numbers or relativity, and it seems very unlikely that anyone will do so for
many years.’ While the view of some that Albert Einstein invented the atomic
bomb is ludicrous, to say on the other hand that there is no warlike purpose for
the equivalence of mass with energy is equally ludicrous. And as for number
theory, much of the work in the field, I am told, is simply classified, because it
could be used, and is used, in cryptography. If somebody came up with a good
algorithm for factoring large numbers he would probably be arrested.
I leave Professor du Sautoy, with the following challenge: can you think of
an explanation why beautiful mathematics tends to be useful in applications?
3
Mathematical reality
John Polkinghorne
Mathematical reality
Considerations relating to the issue of the reality of a noetic world of math-
ematical entities bear some analogy to similar arguments that can be made
defending the reality of the physical world against the critiques of the idealists.
However, before going on to consider these metaphysical arguments, one must
first be clear what results might be expected of them. The character of the
conclusions reached will be insightful and persuasive, rather than logically
coercive. The strict language of ‘proof’, with the implication that only a fool
could disagree, is inappropriate in this field of discourse. No one can force an
intransigent sceptic to give up their position, however arid and implausible it
may be. The solipsist, and the person who maintains that the world and our
memories of it came into being five minutes ago, are both logically invulnera-
ble in their absurdities. The best that can emerge from metaphysical disputation
is an argued claim to have attained the best explanation that is available.
The first of the analogies between human encounters with the physical
world and with the mathematical world relates to the consistency of perception,
30 M E A N I N G I N M AT H E M AT I C S
and the mutual coherence of account, reported by different observers. Connes
summarizes this argument by saying,
What proves [too strong a word!] the reality of the material world, apart from
our brain’s perception of it? Chiefly the coherence of our perceptions and
their permanence. . . . And so it is with mathematical reality: a calculation
carried out in several different ways gives the same result, whether it is done
by one person or by several.
Changeux and Connes (1995, p. 22).
Evolution
A final argument for taking the independent reality of mathematical entities
seriously derives from asking how it might have been that profound mathe-
matical ability arose in the course of hominid evolution. It seems clear enough
that some very modest degree of elementary mathematical understanding—
the ability to count, simple notions of Euclidean geometry, and the capac-
ity to make simple logical associations—would have provided our ances-
tors with valuable evolutionary advantage. But whence has come the human
capacity to go far beyond matters of everyday utility, to attain the ability
32 M E A N I N G I N M AT H E M AT I C S
to conjecture and eventually prove Fermat’s Last Theorem, or to discover
non-commutative geometry? Not only do these powers appear to convey
no direct survival advantage, but they also seem vastly to exceed anything
that might plausibly be considered a fortunate spin-off from such mundane
necessity.
The power of evolutionary explanation depends critically on getting the
environmental factors right, as much as it does on getting the genetic factors
right. If the context within which hominid evolution took place was solely
that of the physico-biological dimensions of reality, as strict neo-Darwinian
orthodoxy supposes, the coming-to-be of human mathematical ability would
seem to be an inexplicable excess. Yet one can take Darwinian explanation
absolutely seriously without having to suppose it to be a totally adequate
account of absolutely everything that has happened. If mathematical entities
constitute an independent realm of reality, then mathematics has always been
‘there’, even before mathematicians emerged. It formed the noetic context
within which that emergence eventually took place. While survival pressures
would favour the initial development of a brain structure that afforded access
to limited arithmetical and geometrical thinking, once that modest degree of
contact had been established with mathematical reality, further new devel-
opmental factors would come into play. The drive to assist physical survival
would be supplemented by the effects of a mental influence that one may
call ‘satisfaction’ (Polkinghorne, 2005, pp. 54–55). Intellectual delight would
then draw our ancestors into an exploration of the noetic world of math-
ematical entities, beguiling them to progress far beyond the modest needs
of everyday practicality. Doubtless the development of mental perceptive
power that this involved was made possible by the epigenetic plasticity of
the human brain, much of whose complex structure derives not from genetic
inheritance, but from response to the shaping influence of experience. Belief
in the reality of mathematics makes intelligible our human ability to be
mathematicians, a capacity that otherwise would have seemed inexplicably
gratuitous.
Unreasonable effectiveness
If mathematical entities are a part of reality, then one might expect that the
ontological realm of their existence is not an isolated domain, disconnected
from all else, but that it has subtle connections with other dimensions of the
real. A very striking example of this happening is provided by the connection
found to exist between theoretical understanding in physics and mathematical
properties. It is an actual technique for discovery in fundamental physics to
seek theories that are formulated in terms of equations possessing the unmis-
takable character of mathematical beauty. This beauty is a rather rarefied form
of aesthetic experience, but it is one that mathematicians can readily recognize
and agree about. It involves qualities such as elegance and economy and the
M AT H E M AT I C A L REALITY 33
property of being ‘deep’; that is to say, extensive and surprisingly fruitful
consequences are found to derive from an apparently simple starting point.
The physicists’ search for beautiful equations is no mere aesthetic indulgence
but a heuristic strategy which has proved its worth time and again in the three-
century history of modern physics. Paul Dirac, one of the founding figures
of quantum theory, made his remarkable discoveries through a lifelong and
highly successful pursuit of mathematical beauty. He once said that it is more
important to have beauty in your equations than to have them fit experiment!
Of course, Dirac did not mean that empirical adequacy was ultimately dispens-
able. No scientist could think that. If you have solved the equations of your new
theory and found that the answers do not appear to agree with experiment, that
is undoubtedly a setback. However, it is not necessarily absolutely fatal. No
doubt you have had to have recourse to some approximation scheme in getting
your solution, and maybe you have just made an inappropriate approximation.
Or maybe the experiments were wrong—we have known that happen more
than once in physics. So there would still be at least a residue of hope. But
if your equations were ugly. . . well, there was no hope. The whole history of
physics was against you.
Dirac’s brother-in-law, Eugene Wigner, who also won a Nobel Prize for
physics, once called this remarkable ability of mathematical beauty to unlock
the secrets of the physical universe its ‘unreasonable effectiveness’. How does
it come about that this apparently abstract subject can illuminate our under-
standing of the structure of the physical world? Why are the beautiful patterns
of pure mathematics, discovered by the mathematicians in their studies, so
often found actually to occur in the structure of the world about us? This is not
the place to pursue that particular issue in detail (I personally look to natural
theology for an answer (Polkinghorne, 1998, ch. 1)). It is sufficient for our
present purpose simply to note the fact, and its implication of a deep mutual
entanglement of the physical and the mathematical. Few doubt the reality of the
physical world; they should be prepared to consider acknowledging a similar
reality of the mathematical world that intertwines with it.
Mathematics also entangles with other dimensions of reality. I wish to take
very seriously human encounter with the realm of beauty. I do not think that
our aesthetic experiences are simply some kind of epiphenomenal froth on top
of what basically is just a physical substrate, but they are a form of access
to yet another dimension of reality. Of course, music involves vibrations in
the air, but its appreciation is not to be reduced to the fourier analysis of
those vibrations. There is a deep mystery about the way that the impact of
packets of sound waves on the eardrum can evoke in us what I believe to be
the valid experience of encounter with a timeless beauty. There is an often
recognised kinship between mathematics and music, expressed not only by the
way that individuals frequently display interests and skills in both, but also by
the patterns that both are found to share, particularly in the case of contrapuntal
music.
34 M E A N I N G I N M AT H E M AT I C S
Another aesthetic experience in which patterns play a vital is exhibited
to us by Islamic art. Marcus du Sautoy (2008, ch.3) has given a fascinating
discussion of the symmetries that underlie the elaborate decoration of the walls
of the Alhambra Palace in Granada. In the nineteenth century, group theorists
were able to show that there are just 17 different basic kinds of symmetry that
can be present in such regular coverings of the plane. All 17 are present in the
decorations of the Alhambra, which was built in the thirteenth and fourteenth
centuries. The Islamic artists involved did not know the group theory, but they
had intuitive access to the mathematical reality that their work expressed.
Conclusion
The criterion for assessing the persuasiveness of a metaphysical position is the
seriousness with which it treats, and the adequacy with which it can contain,
the great swathe of basic human experience which it is seeking to make
intelligible through its insight. Schemes that are made parsimonious simply by
the illegitimate Procrustean strategy of the excision of what, from their point
of view, is inconvenient to take into account, are certainly to be rejected. The
argument of this chapter has sought to show that an approach that seeks to take
the actual character and achievements of mathematics with the due seriousness
that they deserve, is one that is best formulated in a metaphysical context that
acknowledges the reality of a noetic world of mathematical entities.
Comment on John Polkinghorne’s
‘Mathematical reality’
Mary Leng
As Lipton points out, the word ‘best’ needs some clarification here. In
particular, we can distinguish between
the explanation best supported by the evidence, and the explanation that
would provide the most understanding or, in short, between the likeliest and
the loveliest explanation.
Lipton (1991: 2004, p. 59)
Advocating inference to the likeliest explanation is, as Lipton points out, rela-
tively uncontroversial but, sadly, fairly unhelpful—if we had a way of knowing
what scenario was most probable, we would certainly infer that over the alter-
natives, but IBE is surely intended in part as a method for discovering which of
alternative possibilities is more probable. For inference to the best explanation
to be a practical rule of theory choice, we need to build on our account of what
makes an explanation lovely in the sense of providing understanding—perhaps
by invoking theoretical virtues such as simplicity, non-ad-hocness, unifying
power, and so on. Whatever our account of ‘loveliest’ amounts to, Polking-
horne clearly thinks that the noetic realm hypothesis provides the loveliest
explanation of the similarities between mathematical and physical inquiry he
indicates. Lovelier, certainly, than Changeux’s account of mathematical objects
as existing ‘in the neurons and synapses of the mathematician who produces
them’—could those neurons and synapses really contain the rich, surprising,
and universally accessible subject matter of mathematical inquiry? And there
are further phenomena for which Polkinghorne thinks the noetic realm also
provides the loveliest explanation. In particular, Polkinghorne points to cases
of sudden and deep mathematical insights (such as those reported by Poincaré
and Ramanujan); the capacity of human reasoners to go beyond the narrow
range of mathematics that we could expect to give us evolutionary advantage;
and the ‘unreasonable effectiveness’ of mathematics in empirical science. To
take just the last of these, Polkinghorne argues that seeing mathematics as
one dimension of reality renders its effectiveness in finding out about physical
reality unsurprising, since one should expect the mathematical realm to ‘have
subtle connections with other dimensions of the real’. But that two systems of
objects exist cannot by itself be enough to explain why facts about one system
are relevant in finding out about facts about the other: my kitchen exists, and so
does the solar system, but if it turned out I could reliably divine facts about the
38 M E A N I N G I N M AT H E M AT I C S
solar system by reasoning about the contents of my kitchen, one might still find
the effectiveness of this reasoning unreasonable. More needs to be said about
why the mere existence of a mathematical realm should render its effectiveness
reasonable.
There are those who look with suspicion on metaphysical speculation,
on the grounds that it is often unclear what kind of evidence could count
in favour of, or against, any particular metaphysical hypothesis. One of the
great merits of Polkinghorne’s contribution to this volume is that it lays out
in clear and precise terms what would be required of his opponent in this
debate: find a better explanation of the phenomena in question that does not
make use of the noetic realm hypothesis. For those (myself included) who
think alternative explanations can be found, Polkinghorne’s paper presents a
formidable challenge.
Reply to Mary Leng
John Polkinghorne
I am grateful to Mary Leng for her helpful analysis of the arguments I sought
to deploy in defense of mathematical realism. Of course she is right that the
existence of two objects does not itself imply a mutual connection, but the rela-
tionship between mathematics and physics is a deep and apparently intrinsic
(unlike her table and the solar system), and I still maintain this encourages the
thought that each is a part of a greater reality.
This page intentionally left blank
4
Mathematics, the mind, and
the physical world
Roger Penrose
1 This figure first appeared in Penrose (1994), but I have used it frequently elsewhere,
such as in Penrose (2004).
42 M E A N I N G I N M AT H E M AT I C S
Platonic
mathematical
world
Mental
world
1
Physical
world
Fig. 4.1 The three ‘worlds’: the physical, the mental and the mathematical.
many theoretical physicists appear to believe, that there is a deep and precise
underlying dependence of the operation of the physical world on a pre-existing
mathematical order—a mathematical order that appears to have great beauty
and sophistication—which is there to be discovered and not simply imposed
upon Nature as a feature of our gropings towards understanding?
To complete the triad of mysteries connecting these worlds is ‘Mystery 2’,
which is concerned with the relation between physical reality and mentality,
particularly conscious mentality. How does consciousness arise in a world
which seems to be governed by entirely impersonal mathematical operations?
Or is consciousness in some sense primary, its presence being an essential
prerequisite for the very existence of a structure that we could, in any sense,
call a ‘universe’? Is it merely the complication, or perhaps some other sophis-
ticated quality, in the construction of our brains that allows this mysterious
phenomenon of consciousness to come about? And, if so, is this complication
or sophistication to be understood solely in computational terms, as seems to
be a common viewpoint in our present computer age. Or is there some other
essential prerequisite to consciousness, that cannot be understood in purely
computational terms? If the latter, is this ingredient something lying hidden in
M AT H E M AT I C S , THE MIND, AND THE PHYSICAL WORLD 43
the physics that we presently use for our descriptions of the operations of the
world? Or must we search for deeper (mathematical?) theories for a physical
description of consciousness to be possible. Or might we perhaps have to look
even farther afield, to an understanding that lies essentially beyond any kind
of science whatever, as could be the implication of an essentially religious
perspective on these issues?
In this chapter I shall be concerned mainly with Mysteries 0 and 1, as
the topic of the Symposium had to do specifically with mathematics. But it
is my opinion that an adequate discussion of these two mysteries cannot be
completely divorced from some discussion of Mystery 2. I shall argue the
case (by appealing to Gödel incompleteness) that the very fact that our minds
are capable of comprehending sophisticated mathematical arguments—at least
under favourable circumstances—leads us to the conclusion that the operation
of conscious minds cannot be entirely computational and, accordingly, that our
minds cannot be the product of an entirely computational physics. It does not
appear to be the case that the physical laws that we presently understand con-
tain anything that is essentially non-computational (where merely ‘random’ is
not to be considered as ‘essentially non-computational’). The conclusion from
this is that there must be something beyond our present-day physical laws that
is operative in the actions of a conscious mind. It is my own viewpoint that this
is strongly indicative of the actions of a conscious brain being dependent upon
areas of physics (probably at the quantum/classical boundary) that lie outside
the scope of our present-day physical theories; yet that the needed physical
revolution may not itself lie too far beyond what is presently understood.
The issue of Mystery 0 is indeed a closely related matter. Part of the reason
for regarding our access to mathematical truth as being something ‘mysterious’
lies in the nature of our ability to perceive the truth of various particular math-
ematical assertions. As Gödel (and Turing) have demonstrated, if we accept
any particular computationally checkable system of procedures P as providing
valid methods of mathematical proof, then we must equally accept the truth
of some proposition G(P), where the truth of G(P) lies beyond the scope of
the procedures P. Consequently, our methods of ascertaining mathematical
truth cannot be entirely reduced to computational procedures that we accept
as valid. Although various logicians have attached different interpretations to
this conclusion, in my own view, it has the clear implication that there must
be something essential lying outside pure computation that is operative in
conscious understanding. (For further discussions of this point, see Penrose,
1997.) But what really is going on in the activity of a conscious mind when
it becomes convinced of the truth of some mathematical proposition remains
profoundly mysterious.
I maintain, also, that this is strongly indicative of mathematical truth being
something objective (as was, indeed, Gödel’s own view), and is not merely
some ‘game’ based on arbitrary rules that arise out of human culture. Yet,
I am prepared to accept that there could be ‘degrees of Platonism’, where
44 M E A N I N G I N M AT H E M AT I C S
some mathematicians might regard the truth of some proposition P as being
an ‘objective’ matter, whereas another might take the view that the ‘truth’ or
‘falsity’ of P is a matter of opinion, depending upon what ‘man-made’ axiom
system is adopted. I believe that it is clear, from Gödel’s construction, that
certain areas of mathematics are indeed ‘objective’, and therefore have an
existence that is outside ourselves. Such an area would be the truth of what
are called ‘1 -sentences’, these being assertions of the form ‘such-and-such a
computational procedure never terminates’ (where ‘computational procedure’
means ‘Turing-machine action’). An example of a 1 -sentence is the famous
Fermat’s Last Theorem. It seems to me that the truth or falsity of a 1 -sentence
is something entirely objective, and so the truth values of 1 -sentences have
a Platonic reality which is not in doubt (although there may be an element of
subjectivity to the issue of whether or not some particular 1 -sentence has
actually been established).
On the other hand, the objectivity of more sophisticated assertions such
as the truth or otherwise of Cantor’s (generalized) continuum hypothesis is
perhaps more questionable, and it would seem to require a stronger form of
mathematical Platonism to require that all such assertions are true or false in an
absolute sense,2 and not merely dependent upon some particular ‘man-made’
axiom system. The issue of whether or not some particular mathematician is,
or is not, a Platonist usually refers to Platonism in this stronger sense. My
own position is not to be particularly troubled by this issue, as a relatively
weak form of Platonism seems to be perfectly adequate for most arguments of
relevance to physics.
Accepting that, indeed, we need to consider a ‘Platonic mathematical
world’ that is merely large enough to encompass the description of physical
laws, we have an additional case for its ‘existence’, lying beyond anything
that is merely brought into being by human culture or imagination. For the
operations of the physical world are now known to be in accord with ele-
gant mathematical theory to an enormous precision. One particularly striking
example (see, for example, Hartle, 2003) is the double-neutron-star system
PSR 1913+16, which has been under observation, now, for some 30 years,
and the agreement between observation of the timing of its pulsar signals
and Einstein’s general theory of relativity (to something like one hundredth
of a thousandth of a second over that entire period) is phenomenal. This
indicates an extraordinary concurrence between the workings of the natural
world at its most fundamental levels (here the very structure of space and
time) and sophisticated mathematical theory. It makes no sense to me that this
concurrence is merely the result of our trying to fit the observational facts
2 It should be made clear that the results of Gödel and Cohen, showing that the
continuum hypothesis is independent of the standard axioms of set theory, do not
in themselves answer the question of whether or not this hypothesis is true in some
absolute sense; see Cohen (1966).
M AT H E M AT I C S , THE MIND, AND THE PHYSICAL WORLD 45
into some organizational scheme that we can comprehend; the concurrence
between Nature and sophisticated beautiful mathematics is something that is
“out there” and has been so since times far earlier than the dawn of humanity,
or of any other conscious entities that could have inhabited the universe as we
know it.3
Acknowledgement
I am grateful to NSF for support under grant PHY 00-90091.
3 More details about the arguments given here are to be found in Penrose (2011).
Comment on Roger Penrose’s
‘Mathematics, the mind, and
the physical world’
Gödel’s theorems and Platonism
Michael Detlefsen
Roger Penrose’s chapter contains a number of claims and ideas that warrant,
and have received, extensive discussion. In this note, I will focus on the
following two claims that are central to his view of the significance of Gödel’s
theorems.
I: Gödel’s incompleteness theorem(s) demonstrate that ‘if we accept any
particular computationally checkable system of procedures P as providing
valid methods of mathematical proof, then we must equally accept the truth
of some proposition G(P), where the truth of G(P) lies beyond the scope
of the procedures of P.’
II: Claim I ‘has the clear implication that there must be something essen-
tial lying outside pure computation that is operative in conscious
understanding.’
We can restate Penrose’s claims more clearly and in more familiar
terminology as follows:
I∗ : For any formal system P, if we accept all of P’s axioms as true, and all its
rules of inference as valid, we are rationally obligated to accept P’s Gödel
sentence G(P) (and its P-equivalent consistency formulae Con(P)) as true
too.
II∗ : Claim I∗ clearly implies that there is a set A of sentences we are rationally
obliged to accept that cannot be formalized (i.e., is not a computably
enumerable set).
There is little reason I can see for confidence in either I∗ or II∗ . G(P) and
the usual consistency formulae Con(P) equivalent in P to it are not logically
M AT H E M AT I C S , T H E MIND, AND THE PHYSICAL WORLD: C O M M E N T 47
implied by P. This means that they logically imply sentences that are not
logically implied by P. There is generally speaking no reason to believe that
all reasoning which supports P will equally support these ‘extra’ implications.
Consequently, there is no general reason to think that rational acceptance of P
will rationally oblige acceptance of G(P)/Con(P).
This may seem wrong to those who believe that evidence capable of
justifying P must also be capable of justifying belief in its consistency. It is
important to realize, however, that to justify the consistency of P and to justify
G(P) or Con(P) are not the same thing. Justification that P is consistent is not
in and of itself justification of G(P) or Con(P). For the latter, justification for
the following supplementary proposition is also required.
Supp: If P is consistent, then G(P)/Con(P).
Such justification will not, however, necessarily be included in the evidence
one might have for P’s consistency.
This is not to deny, of course, that there may be justification for Supp. Nor
is it to deny that there may be rationally compelling grounds for P that include
rationally compelling grounds for G(P)/Con(P). It is intended only to indicate
the non-triviality of Supp and thus to demur to claims such as I∗ , which
suggest that rational acceptance of G(P)/Con(P) is somehow presupposed by
rational acceptance of P.
There are other reasons too to question Penrose’s argument, but there is no
space to enter into these here.
This page intentionally left blank
5
Mathematical understanding
Peter Lipton
He then briefly suggests that we shift from thinking about the context of
justification, and focus on the context of discovery.
M AT H E M AT I C A L U N D E R S TA N D I N G : A D D E N D U M 59
On the broadly Fregean picture sketched just above, however, we may not
have to leave the realm of justification in order to find a role for inference to the
best explanation in mathematics. The Lipton–Rosen–Frege idea is that some
mathematical propositions rest on, or depend on, others. As Frege was aware,
the regress cannot go on forever. Some propositions lie at the base of the foun-
dational enterprise, and are not explained by anything. These are the axioms.
Earlier in the essay, Lipton suggests that when it comes to proper axioms, there
are no legitimate ‘why-questions’, or at least no legitimate answers to ‘why-
questions’: ‘Thus one might hold that mathematical understanding flows from
the special epistemic status of mathematical axioms, and that these axioms are
where the regress stops.’
But how are the axioms known? The traditional view is that the axioms
are ‘self-evident’. A full and complete understanding of an axiom immediately
gives rise to justification for it. It seems to me, however, that this, traditional
foundationalist view is not really tenable when it comes to modern mathemat-
ics. Some propositions presented as axioms are hardly obvious, and it is a bit of
a stretch to say that a full and complete understanding of them will justify them
(see Shapiro, 2009). Perhaps we can say, instead, that at least some axioms
are chosen, not because of any intrinsic or self-evidence they may have, but
because they make for a good, or, as Lipton might put it, lovely explanation of
some of the theorems. It is a holistic picture. In a well-systematized branch of
mathematics, the theorems are shown to depend on the axioms. In Aristotelian
terms, the ‘why’ of the theorems lies ultimately in the axioms. When we turn
to the axioms, and ask why they are true, or at least how it is that we know
them, the answer is that they provide the best explanation of the theorems.
This page intentionally left blank
6
Creation and discovery
in mathematics
Mary Leng
One important role for the philosophy of mathematics is to account for the
phenomenology of the discipline, that is, to account for what it feels like to do
mathematics. One aspect of this phenomenology is the sense mathematicians
often have that they are discovering, rather than creating or inventing, the
nature of mathematical reality. Given this aspect of mathematical practice, a
natural assumption is that mathematicians are involved in the investigation
of a mathematical reality that is independent of creative human decisions,
and independent of our beliefs about that reality, in much the same way that
physical scientists are involved in the investigation of a physical reality whose
nature does not depend on us. If we accept this assumption, and the analogy
on which it is based, then the question arises: ‘What is the nature of this
mathematical reality, and how is it possible for us to have knowledge of it?’
Taking seriously the analogy with physical science would suggest that
mathematicians investigate a realm of mathematical objects, inquiring into the
nature of numbers, sets, etc., just as physical scientists inquire into the nature
of atoms. But if there is a mathematical realm of independently existing, non-
physical objects, over and above the realm of physical objects we ourselves
inhabit, then the question of how mathematical knowledge is possible becomes
pressing. Our knowledge of the physical realm stems from our interaction,
as physical beings, with that realm, but can mathematical knowledge be
accounted for in an analogous way? Mathematician G. H. Hardy described
mathematical discovery as observation of mathematical reality:
I believe that mathematical reality lies outside us, that our function is to
discover or observe it, and that the theorems which we prove, and which
we describe grandiloquently as our “creations,” are simply the notes of our
observations.
(Hardy, 1940)
62 M E A N I N G I N M AT H E M AT I C S
But is this account of mathematical discovery as rooted in observation of a
mathematical realm tenable?
Certainly, Plato thought that knowledge of the mathematical realm could
be accounted for as resulting from a kind of direct ‘observation’. According
to Plato, mathematical objects belong to an eternal realm of Forms, which
are directly perceived by the immortal, immaterial soul prior to its ‘birth’ as
a flesh-and-blood human. In our physical incarnations, what mathematical
knowledge we have is had by recollection of this direct experience of
the Forms (Plato, Meno, 81d–86c). Our theorems then report what our
mathematical inquiry has enabled us to remember about our earlier direct
observation of the mathematical realm. However, many would find this
account of our knowledge of mathematical reality hard to swallow, requiring
as it does some powerful and potentially problematic assumptions about mind
and body. Perhaps something like Plato’s picture could be defended on the
basis of an inference to the best explanation: implausible as it may sound,
it must ultimately be accepted as the only good way of accounting for the
phenomenology of mathematical discovery. However, before opting for this
solution, it is worth examining the phenomena in question, to consider what
alternative explanations might be available.
Just what is it that mathematicians seem to discover? The discoveries
Hardy mentioned were of theorems, presumably within established math-
ematical theories. But mathematicians also, of course, create/discover new
mathematical theories, within which theorems can be proved. We must, then,
consider whether either of these kinds of discoveries are best viewed as dis-
coveries of the nature of a realm of mind-independent, non-spatiotemporal
mathematical objects.
One might think that the discovery of entirely new theories provides the
best evidence for such a mathematical realm. After all, once the axioms or
basic presuppositions of a theory are in place, our discovery of mathematical
theorems is discovery of what follows from those presuppositions, and at least
at first glance, this kind of ‘what if’ inquiry, into what would have to be true
if our mathematical axioms were true, does not require that our mathematical
axioms are in fact true of some underlying mathematical reality (though more
on this later). On the other hand, in developing new mathematical theories,
mathematicians often have a sense of discovering the basic assumptions of
these theories, as assumptions that truly describe an important corner of mathe-
matical reality, rather than simply plucking theoretical hypotheses out of the air
with a view to inquiring blithely into their consequences. Surely this aspect of
the phenomenology of mathematical practice provides the strongest evidence
for an independently existing realm of mathematical objects?
In fact, I will argue, the phenomenology of theory development is actu-
ally easier to account for from an anti-Platonist perspective than is the phe-
nomenology of mathematical proof within theories. If accounting for the
phenomenology of mathematical discovery requires us to posit any kind of
‘reality’ to ground our mathematical judgements, this reality is not a realm
of mathematical objects, but rather, I claim, a realm of objective facts about
C R E AT I O N A N D D I S C O V E R Y I N M AT H E M AT I C S 63
logical consequence. Insofar as we are concerned with understanding the sense
of discovery that is present in mathematical theorizing, the real puzzle to be
accounted for (a puzzle that, in fact, already arises even when one considers
ordinary empirical reasoning) concerns what Wittgenstein called ‘the hardness
of the logical must’ (Wittgenstein, 1953: 2001, I, §437), and not the existence
of a realm of mathematical objects.1
Let us start, though, with theory development. It is certainly true that the
choice of assumptions for a new mathematical theory is usually far from arbi-
trary, and indeed that the development of appropriate theoretical assumptions
is often rightly viewed as a significant and extremely nontrivial achievement.
But does this require us to view the development of new mathematical the-
ories as the description of an independently existing realm of mathematical
objects?
The evidence of mathematical practice, I think, speaks against this, sug-
gesting as it does constraints on our theory development that account for
our sense of discovery without requiring us to posit a realm of mathematical
objects to be discovered. For, very often, mathematical theories are developed
as solutions to problems we have set ourselves, where the constraints of the
problems are enough to narrow down the range of options that could count
as an appropriate solution (often even pinning down a unique solution). Take,
for example, W. R. Hamilton’s discovery of the quaternions which, he tells us,
‘started into life, or light, full grown, on the 16th of October, 1843’ (quoted
in Tait, 1866, p. 57). Was Hamilton’s moment of inspiration, the discovery of
the equation i2 = j2 = k2 = −1, which he excitedly carved into the stone of
Brougham Bridge, a sudden recollection of a truth contemplated by his pre-
embodied soul?
In fact, as Hamilton’s own description of his 15-year struggle to develop
rules of addition and multiplication for a three-dimensional analogue of two-
dimensional complex numbers shows, his moment of inspiration is better
viewed as a sudden realization of what had to hold, given the constraints he
had set himself. Hamilton’s aim was to discover laws for multiplying triplets
of the form x + iy + jz, along the lines of the laws for multiplying pairs x + iy,
where j was to be viewed as a square root of −1 distinct from i. A constraint he
set himself was to satisfy the ‘law of the moduli’: the modulus of the product of
two triplets should equal the product of the moduli of the two triplets taken sep-
arately. That is, if (a + ib + jc) · (x + iy + jz) = u + iv + jw, then the law of
the moduli would require that (a2 + b2 + c2 ) · (x2 + y2 + z2 ) = u2 + v2 + w2 .
This constraint, it turns out, is impossible to satisfy if multiplication is assumed
1 This is not to say that there are no aspects of mathematical practice that might
require us to assume the existence of mathematical objects. Indeed, if we turn to the
question of the applicability of mathematics, considerations of scientific confirmation
might require us to hold, not only that there are objective facts about the conse-
quences of our hypotheses concerning mathematical objects, but also that some of those
hypotheses are in fact true.
64 M E A N I N G I N M AT H E M AT I C S
to have its usual properties of commutativity and associativity. Dropping the
commutativity constraint, however, showed Hamilton that there was room for
a solution in some special cases, so long as the product ij was equal to −ji,
and that neither were equal to 0. This effectively forced Hamilton’s hand,
leading him to discover that the only possible value for the product ij was a
third imaginary, k.
The exact problem Hamilton originally set himself had no solution. How-
ever, the system of quaternions emerged as the best way to fit as many of
Hamilton’s original constraints as possible. In fact, if we set the problem in
advance as that of extending multiplication to a system of numbers based on n
square roots of −1, preserving associativity and the law of the moduli, it is a
matter of logical consequence, rather than a mathematical ‘matter of fact’, that
only three solutions are possible, these being n = 0 (the real numbers); n = 1
(the complex numbers), and n = 3 (the quaternions).
In more pedestrian cases of mathematical theory development than Hamil-
ton’s, we often find that axiomatizations are governed by the constraint to
come up with a system of assumptions that will pin down essential features of
an already familiar mathematical or empirical system. For example, Euclid’s
axioms for geometry were intended to pin down truths about points and straight
lines in physical space, and were ‘discovered’ through examination of the ques-
tion of what had to be assumed in order to prove many other results believed
to be true of points, lines and geometric shapes. And aside from physical
interpretations, the development of axioms for mathematically familiar objects
is also commonplace—as, for example, in the development of the Dedekind-
Peano axioms, where the axiomatization is constrained by the requirement that
the structure axiomatized be an ω-sequence. Indeed, axiomatizations of mathe-
matically familiar objects often appear initially as theorems—for example, the
axioms of my own favourite corner of ‘mathematical reality’, C∗ -Algebras,
first appeared as a part of the Gelfand-Naimark theorem, which showed that
those axioms pinned down up to isomorphism subalgebras of the algebra B(H)
of bounded operators on a Hilbert space, structures which were already of
independent mathematical interest.
We are constrained, in all of these cases, by what is known already about
the system to be captured, and this constraint of course leads to the sense
that, in coming up with a formal theory, we are ‘getting something right’.
In such cases, our axioms might strike us as being ‘true’, rather than merely
convenient or interesting, but this sense of correctness is explicable even if we
do not invoke the existence of a realm of objects for our axioms to be true
of. What counts as getting something right in these cases is dependent on the
constraints we have set ourselves (which require us to pin down a structure
satisfying various assumptions). We need not (though we may) think of these
constraints as being imposed by an independent realm of objects about which
our theories must assert truths. Rather, as consequences of a collection of initial
assumptions or constraints, new axiomatizations can come about in much the
way that theorems do within the context of already established theories. In each
C R E AT I O N A N D D I S C O V E R Y I N M AT H E M AT I C S 65
case, the area of mathematical discovery that really matters seems to be the
discovery of consequences of one’s assumptions.
Let us turn, then, to the kind of mathematical theorizing that takes place
against a backdrop of accepted assumptions (such as axioms). In the paradigm
cases, such reasoning is deductive, and amounts to the proving of theorems
from axioms, although there is also some room for the use of abductive reason-
ing in such contexts: mathematicians may reason that, given their assumptions,
such and such a result is likely to be true. Sticking for now with the central
case of deductive proof from axioms, we can consider what conclusions can
be drawn from the sense mathematicians often have that in such activity they
are involved in discovery rather than creation of mathematical results. Are
mathematicians who are engaged in proving theorems really discovering the
already determined consequences of their assumptions, or could it be the
case that, despite the strong sense of discovery, they are actually involved in
the creation of links between axioms and theorems that were not already, in
some sense, ‘out there’? If mathematicians are engaged in discovery rather
than creation, what implications does this have for our view of the nature
of mathematics. In particular, is this discovery discovery of a realm of mind-
independent mathematical objects? On the other hand, if we do choose to view
deductive proof from axioms as a matter of creation, rather than discovery, can
this be reconciled with the felt objectivity of mathematical proof, and, indeed,
the applicability of mathematical reasoning?
A natural line of thought takes it that, yes, deductive mathematical rea-
soning is objective, leading us to the discovery of logical consequences of
our mathematical assumptions. But such objectivity has nothing to do with
an independent realm of mathematical objects, but is entirely a result of the
objectivity of logic. After all, in reasoning to a theorem P on the basis of math-
ematical assumptions A1 , . . . An , we prove not that P is true, but rather, that
if A1 & . . . & An , then P. This conditional claim does not assert the existence
of any mathematical objects; its truth, we may suppose, rests solely on the
fact that P is a logical consequence of A1 , . . . , An . According to this way of
thinking, there is nothing particularly problematic about the felt objectivity of
mathematical proof from axioms—it is simply a special case of the objec-
tivity of any deductive reasoning, and depends solely on the objectivity of
the ‘following from’ relation. Furthermore, as we have seen, a similar story
can be told about the kind of reasoning that leads to the development of new
mathematical theories: although this reasoning does not start from axioms, it is
nevertheless governed by logical constraints established by the consequences
of our preformal mathematical assumptions and/or desiderata.
There is, however, a problem with this comfortable-seeming position,
which arises once we consider what we mean by the claim that P follows
logically from A1 , . . . , An . We surely do not mean by this that P can be
derived from A1 , . . . , An using an accepted collection of rules of inference. For
one thing, we know of cases where this analysis falls short of capturing our
usual notion of logical consequence: take the second-order Peano axioms for
66 M E A N I N G I N M AT H E M AT I C S
arithmetic. Gödel’s first incompleteness theorem tells us that, for any (stan-
dard) collection of inference rules we can come up with, there will be a
sentence G in the language of (2nd order) Peano arithmetic which follows
logically from those axioms but which is not derivable using our chosen
collection of rules. But even without the worries that Gödel’s theorem brings,
one should be wary of an analysis of logical consequence which rests facts
about logical consequence on a chosen collection of inference rules; after all,
what makes a collection of inference rules a good one is, presumably, that those
rules respect facts about logical consequence, and not vice versa.
All of which is simply to say that the relevant notion of logical consequence
underlying the objectivity of mathematical reasoning is that of semantic, rather
than syntactic consequence. In the relevant, semantic, sense, P is a logical
consequence of A1 , . . . , An if and only if it is not logically possible for P to
be false while A1 , . . . , An are all true. But this analysis simply replaces one
undefined logical notion (logical consequence) with another (logical possibil-
ity). If our question is what grounds the objectivity of logical consequence,
then surely an analogous question arises for logical possibility: can we say
anything more about what the logical possibility or impossibility of a sentence
amounts to?
Here is where the difficulty for our comfortable view of the objectivity
of mathematical reasoning arises. For, arguably, the best analysis of logical
possibility available is mathematical: a sentence P is logically possible if there
is a set theoretic model in which that sentence is interpreted as a truth. On this
analysis, P is a logical consequence of A1 , . . . , An if and only if, in all models
which make A1 , . . . , An true, P is also true. If this analysis is correct, then
the existence of objective facts concerning logical consequence comes down
to the existence of a realm of mathematical objects (set theoretic models).
So the objectivity of mathematical discovery is, after all, dependent on an
objective realm of mathematical objects, and we are led back to the difficulty
of explaining our knowledge of such things.2
Is there an alternative analysis of logical possibility available? One might
try to eschew abstract mathematical objects in favour of logically possible
concrete worlds. But even if we could make sense of the notion of a logically
possible world in such a way as to respect our intuitions concerning logical
possibility, if it is facts about these worlds that ground the objectivity of math-
ematical inference, we will still have difficulty explaining how we could have
knowledge of the following-from relation, since such worlds are presumably
spatiotemporally isolated from our own. Since we do seem to know some
facts about what follows from what, we should be wary of any analysis of the
the proof changes the grammar of our language, changes our concepts.
It makes new connections, and it creates the concept of those connections.
(It does not establish that they are there; they do not exist until it makes them.)
(Wittgenstein, 1956: 1978, III, p. 31)
Perhaps, then, despite appearances, there are no objective facts about logical
consequence in mathematics, just the results of human decisions that could
always have gone differently?
68 M E A N I N G I N M AT H E M AT I C S
Perhaps this view could be sustainable if individual mathematical theories
were entirely isolated from one another, so that ‘decisions’ made within one
theory would not clash with decisions made elsewhere. Indeed, Wittgenstein
himself thinks that cross-theoretical links are themselves a matter of decision,
that, for example, it is a matter of choice to embed the natural numbers in the
integers and so on (see, e.g., Waismann, 1979, pp. 34–6). But here the phenom-
enology of mathematical discovery speaks strongly against the conventionalist
position, as Friedrich Waismann notes in explaining (in his (1982) paper,
‘Discovering, Creating, Inventing’) his own abandonment of Wittgensteinian
conventionalism. There are just too many examples of theorems proved in
one mathematical context bearing out (and even illuminating) the conclusions
drawn in other areas. Waismann’s example is of a result concerning the real
numbers that receives an explanation once the reals are embedded in the
complex numbers. The Taylor series expansion
1
= 1 − x2 + x4 − x6 + · · ·
1 + x2
converges for |x| < 1, but diverges for all other real values of x. Once we
1
embed the reals in the complex numbers, the behaviour of the function 1+z 2 in
its real portion is explained by the fact that the complex function has singular-
ities at z = ±i, together with a theorem of complex analysis which tells us that
any power series expansion only converges within a circle of radius R about the
origin, and diverges elsewhere. Given these facts about complex functions, the
real-valued function could not have behaved other than it did. Far from being
a matter of human convention, such results seem determined independently of
our choices. It is, as Waismann remarked, as if the real function already knew
that the complex numbers were there.
Related to this issue (which we might call the applicability of mathematics
within mathematics), is the phenomenon of the applicability of mathematics
to nonmathematical questions. We are abundantly aware of cases where math-
ematical reasoning is used to derive empirical predictions, and where these
predictions turn out to be correct. One view on the applicability of mathematics
takes the applicability of mathematics to reside in structural similarities: a
mathematical theory is (sometimes) applicable to a nonmathematical phe-
nomenon because the nonmathematical reality is structurally similar to some
portion of the structure described by the mathematical theory in question. But
if in mathematical reasoning we were simply freely deciding, at each step,
what we will take to be true of the objects of a given mathematical theory,
it is surely entirely mysterious how these free decisions so regularly result in
accurate predictions.
Both of these phenomena, then, speak against the radical anti-objectivist
account of mathematical reasoning. What, then, does the phenomenon of
mathematical discovery have to tell us about mathematical reality? Not, I think,
that our mathematical theorems are answerable to an independent realm of
C R E AT I O N A N D D I S C O V E R Y I N M AT H E M AT I C S 69
mathematical objects. But if we do not accept Wittgenstein’s extreme conven-
tionalism, at the very least we must accept that our mathematical discoveries
are underpinned by objective facts about logical consequence. And if we wish
to hold, with Kreisel, that the problem that concerns us is ultimately ‘not
the existence of mathematical objects, but the objectivity of mathematical
statements’ (Dummett, 1978, p. xxviii), then we will have to accept that the
relevant facts concerning logical consequence do not reduce to facts about
mathematical models. If we want to understand mathematical discovery, then,
we must consider from where the objectivity of these facts might arise.
Comment on Mary Leng’s ‘Creation
and discovery in mathematics’
Sensing objectivity
Michael Detlefsen
Introduction
This chapter is an investigation into the question whether there are features
of our acquisition of mathematical knowledge that support a realist attitude
towards mathematics. More particularly, it is a reflection on the reasoning
which moves from the claim that
I. mathematicians are commonly convinced that their reasoning is part of
a process of discovery, and not mere invention,
to the claim that
II. mathematical entities exist in a noetic realm to which the human mind
has access.
For convenience, I’ll refer to this as the original argument.
The use of the term ‘noetic’ in II calls for brief comment. Traditionally it
has been used to signify a type of apprehension, noēsis, which is characterized
by its distinctly ‘intellectual’ nature. This has generally been contrasted to
forms of aisthēsis, which is broadly sensuous or ‘experiential’ cognition, or
intuition. There is interest, and difficulty, in determining more exactly the ways
in which an intellectual experience of a supposedly non-sensuous reality might
resemble and might also contrast with sensuous experience of material objects.
This is where a great deal of the difficulty concerning the content of ‘noetic’
will be met. The terms of such a comparison will therefore be one of our chief
concerns.
74 M E A N I N G I N M AT H E M AT I C S
Experience and involuntariness: background
For Plato, the objects of noēsis were the Forms, which he believed to be
manifested by experience while also transcending it. The empiricists, on the
other hand, generally associated intellectual apprehension with apprehension
of concepts or ideas. These, when legitimate, were mental representations
obtained by abstraction from sensory experience.
Kant famously emphasized a distinction between two types of representa-
tions, intuitions (Anschauungen) and concepts (Begriffe). Among their impor-
tant differences, he maintained, was one concerning the extent to which they
are within the power of the judging agent to control. Concepts were taken to
be spontaneous (see Kant, 1781: 1990, A50–51/B74–75), or capable of being
brought into existence by a judging agent’s own intellectual initiative (selbst
ausgedachten, op. cit., A639/B667). Intuitions were not. In the end, however,
Kant required that genuine or legitimate concepts be consistent (nicht selbst
widersprechen, op. cit., A150/B189). The freedom to create or generate them
was thus a constrained freedom.
This notwithstanding, Kants saw a great difference between our control
over concepts and our control over intuitions (see op. cit., A19/B33, B132).
He regarded intuitions and their relations as given (gegeben) (see op. cit.,
A19/B33, B132) and not under the spontaneous productive control of our
minds. Concepts, on the other hand, could be thus produced. There was there-
fore no guarantee that they be exhibited by any object(s).
. . . even if our judgment contains no contradiction, it may connect concepts in
a manner not borne out by any object, or in a manner for which no ground is
given . . . and so may still, in spite of being free from all inner contradictions,
be either false or groundless.
Kant (1781: 1990, A150; B190)1
Kant’s acceptance of this asymmetry between concepts and intuitions provides
an interesting point of contrast to more recent views of the nature of concepts.
Specifically, it seems to be sharply at odds with the view of concepts presented
by Gödel in various foundational writings of the 40s, 50s and 60s (see Gödel,
1947: 1990). It is to Gödel’s view(s) that I now turn.
Gödel agreed with Kant in regarding the involuntariness of a representation
as a mark of its objectivity. However, whereas Kant regarded our use of
mathematical concepts as essentially creative or voluntary,2 Gödel regarded
it as distinctly involuntary. Contrary to Kant, he thus maintained that
. . . despite their remoteness from sense experience we do have something like
a perception also of the objects of set theory, as is seen from the fact that the
. . . I can think (denken) whatever I want, provided only that I do not con-
tradict myself, that is, provided my concept (Begriff) is a possible thought
D I S C O V E R Y, INVENTION AND REALISM 75
axioms force themselves upon us as being true. I don’t see any reason why we
should have less confidence in this kind of perception, i.e., in mathematical
intuition, than in sense perception.
Gödel (1947: 1990, p. 268)
Indeed Gödel claimed more than the involuntariness of our apprehension
of concepts. He believed that it has a perception-like, but still non-sensory
character. He maintained as well that the cognitions it yields are in important
ways independent of both the voluntary acts and involuntary dispositions of
our minds.
I am under the impression that . . . the Platonistic view is the only one tenable.
Thereby I mean the view that mathematics describes a non-sensual reality,
which exists independently both of the acts and the dispositions of the human
mind and is only perceived, and probably perceived very incompletely, by the
human mind.
Gödel (1951: 1995, pp. 322–23)
Despite these realist convictions, however, Gödel conceded certain points to
the conventionalist. Specifically, he believed, they were right to think that
mathematics is about concepts rather than physical or psychical items (see
Gödel, 1951: 1995, p. 320). They were right too, he said, to believe that
mathematical truths in some sense owe their truth to the meanings of terms—
specifically, to the concepts expressed by terms (loc. cit.). Where the con-
ventionalist went wrong, he believed, was in taking these meanings to be
determined by conventions (ibid.). The truth as he saw it was rather that
these concepts form an objective reality of their own, which we cannot create
or change, but only perceive and describe.
Gödel (1951: 1995, p. 320)
In Gödel’s view, then, mathematical concepts are discovered and not cre-
ated by acts of convention or other mental acts or dispositions. Similarly for
truths concerning them (loc. cit.). The chief evidence of this, in his view,
was the involuntariness with which truths concerning mathematical concepts
‘force’ themselves on us as being true. This, to Gödel, was signal indication
of their independence from our creative capacity, a capacity he took to mainly
(möglicher Gedanke). This suffices for the possibility of the concept (Begriff),
even though I may not be able to answer for there being, in the sum of all
possibilities, an object (Objekt) corresponding to it. Indeed, something more
is required before I can ascribe to such a concept objective validity (objektive
Gültigkeit), that is, real possibility (reale Möglichkeit); the former possibility
is merely logical. This something more need not be sought in the theoretical
sources of knowledge (theoretischen Erkenntnisquellen); it may lie in those
that are practical.
Kant (1781: 1990, Bxxvi, note *)
76 M E A N I N G I N M AT H E M AT I C S
be constituted by our mental dispositions and our various abilities to perform
voluntary mental acts.
5 Gödel’s exact statement was ‘the axioms [of set theory] force themselves upon us
as being true.’ ‘The’ axioms? Was Gödel assuming that there is a unique, or perhaps
uniquely best, axiomatization of set theory? Not necessarily. Given a set-theoretic
language L, there is no necessary incoherence in believing that (i) there is an identifiable
set of propositions formulable in L whose evidentness does not derive from that of
other propositions formulable in L, and also that (ii) not all elements of need or
even ought to be taken as axioms of an axiomatization of set theory. There might, for
example, be logical overlap between elements of that would make it unnecessary or
even undesirable to take all of them as axioms of an axiomatization of set theory. There
might also be alternative ways of thinking about sets, or different concepts of set, that
would divide the elements of in such ways as would associate certain elements of
with certain concepts or ways of thinking and not others.
78 M E A N I N G I N M AT H E M AT I C S
Such reasoning, of course, raises many questions. Among these are:
Q1: How reliable and revealing an indicator of external reality is our expe-
rience of the forcedness of sensory propositions?
Q2: How reliable and extensive is the asserted analogy between the forced-
ness of basic mathematical propositions and the forcedness of basic
sensory propositions as indicators of external realities?6
Q1 queries the evidential connection between the property of sensory
judgements called forcedness and the existence of an external source respon-
sible for it. Affirmation of such a connection would seem to require belief that
forcedness of sensory judgement is best explained by some sort of transfer
of energy—specifically, transfer to a sensory agent from a sense-stimulative
source external to her.
Understood this way, is phenomenal forcedness of sensory judgements a
reliable indicator of external reality? Here, I think, we need to distinguish
two types of such reliability. One is what I’ll call existential reliability, or
reliability concerning the existence of an external source for an experience
of forcedness. The other is what I’ll call attributive reliability, or reliability
concerning the characteristics of a supposed external source forced on us in a
sensory judgement.
Exceptions to both existential and attributive reliability are, of course,
familiar from the literature on sense perception and I’ll not go into them in
any detail here. Rather, I’ll simply note that quasi-sensory experiences such as
certain hallucinations provide cases where an experience of forcedness is not
an existentially reliable indicator of an external reality. Similarly, well-known
cases of optical illusion raise similar concerns regarding attributive reliability.
In addition to these concerns, there are three others I’ll mention. The first
has to do with our understanding of ‘forcedness’. What did Gödel mean when
he said that our mathematical experience includes the experience of proposi-
tions’ forcing themselves upon us as being true? A natural interpretation would
include the following implication: P’s being forced upon us as true implies that
we form a belief that P.
This raises an important question concerning Gödel’s supposed mathemat-
ical ‘perception’. The reason why is that sense perception does not seem to
sponsor the above implication. That is, having a sensory experience as of P
(e.g., an experience as of one line segment’s being longer than another) does
not seem to imply that we form a belief that P. There are well-known illusions
(e.g., the Ponzo and Müller-Lyer illusions) in which we experience one line
segment’s being longer than another but do not believe it to be so.
6 More basic than either Q1 or Q2, of course, is the difficult question of how to make
sense of the notion of a sensory proposition. I won’t address this problem here, though,
since it is Q1 and Q2 and their focus on the phenomenon of forcedness that are my
chief concerns.
D I S C O V E R Y, INVENTION AND REALISM 79
In sensory perception, then, we can have an experience as of P but not
judge or believe that P. Sensory perception seems to be a form of apprehension
in which not all contents it presents are presented as true. Is the same true of
Gödel’s mathematical ‘perception’? That is, can mathematical perception in
some sense ‘give’ appearances which, despite there being given in this way,
are nonetheless not forced on us as being true?
I don’t know the answer to this question, but it’s not the answer that is
my chief concern here. Rather, it is the question itself. It shows, if I am not
mistaken, a certain type of difficulty posed by Q2, a difficulty that will confront
any attempt such as Gödel’s to exploit an analogy between sense perception
and a perception-like form of apprehension in mathematics.
Nor is this the only such difficulty or the most serious. More troublesome,
I think, are problems raised by cases of sensory illusion where false contents
are forced on us as being true. A well-known illustration is Adelbert Ames’
‘Distorted Room’. Viewed in such a way as to eliminate stereoptic information
(e.g., viewed through a peephole), the room looks like an ordinary ‘cubical’
room with rectangular windows, a flat, rectangular, level floor and rectangular
walls of equal and uniform height and depth.
When commonplace objects (e.g., an adult human being of normal size and
a smaller child of normal size) are placed on opposite sides of the room and
photographed, and the photographs are merged into a single image, strange
appearances occur. The child, for example, will appear to be much larger than
the adult.
The truth, of course, is that Ames’ room is not, despite its appearance, an
ordinary cubical room. Its seemingly rectangular windows, walls and floor are
trapezoidal, its walls are not of uniform height and depth and the floor is not
level. That it looks normal, and continues to look normal when familiar types
of objects are placed in it suggests the influence that dispositions can have on
perception.7 As R. L. Gregory described it:
Evidently we are so used to rectangular rooms that we accept it as axiomatic
that it is the objects [human bodies] which are odd sizes rather than that the
room is an odd shape. But this is essentially a betting situation—it could be
either, or both, which are peculiar. Here the brain makes the wrong bet, for the
experimenter has rigged the odds. Indeed . . . the most interesting feature of
the Ames Distorted Room is its implication that perception is a matter of mak-
ing the best bet on the available evidence . . . wives do not see their husbands
as distorted by the Room—they see their husbands as normal, and the room
its true queer shape . . . [Also] Familiarity with the room, especially through
touching its walls . . . does gradually reduce its distorting effect on other
objects, and finally it comes to look more or less as it is—distorted in fact.
Gregory (1969, pp. 180–181, brackets added)
For more on the Principle of Permanence see Detlefsen (2005, pp. 273–278).
D I S C O V E R Y, INVENTION AND REALISM 81
knowing its nature and origins. Frege had a particularly desultory view of the
conditioning that he took to lie behind many of our affirmations.9
One need only use a word or symbol often enough, and the impression will
be produced that this proper name stands for something; and this impression
will grow so strong in the course of time that in the end hardly anybody has
any doubt about the matter.
Frege (1903: 1962, Vol. II, §142)
Neither Gödel nor others since his time have adequately addressed these
difficult problems. And though this should not be taken to suggest that they
are irresolvable, it should motivate consideration of alternative approaches to
Gödel’s seeming understanding of the invention/discovery question in mathe-
matics. It is to the consideration of such alternatives that I now turn.
Traditional alternatives
Gödel’s view shares a concern with the more important traditional views con-
cerning the discovery vs. invention issue. Broadly speaking, this is a concern
for the objectivity of mathematics. A little more exactly, it is a concern to
restrict certain types of subjectivity in mathematics, specifically those having
to do with the possibility of legitimately using ‘created’ or ‘invented’ concepts.
I’ll now briefly survey some of the more interesting parts of the historical
literature that touch on this issue.
Ancient views
The extent to which mathematics should be open to creation or invention
was a common concern in antiquity. Looking at the ancient literature, though,
we quickly notice a terminological difference. Specifically, what is currently
meant by ‘discovery’ is quite different from what the ancients meant.
This can be seen from Cicero’s (106 BC–43 BC) use of the term. He
distinguished two types of methods proper to careful, systematic inquiry in
any area of thought, mathematics included. Methods of the one type he called
methods of ‘discovering’ (Cicero, 1894–1903, p. 459), methods of the other
type, methods of ‘deciding’ (ibid.). In time this grew into the traditional
division between artis inveniendi (arts of invention, arts of discovery or arts of
investigation) and artis iudicandi (arts of adjudication, also sometimes called
artis demonstrandi (arts of demonstration)).
This distinction was important in science and also in jurisprudence.
In mathematics, it generally took the form of a distinction between
9 In the concluding section I’ll discuss two others, Bolzano and Dedekind, who
also emphasized the elaborate conditioning that underlies even ordinary mathematical
beliefs.
82 M E A N I N G I N M AT H E M AT I C S
discovermental and demonstrative methods of investigation. The end of the
former was the efficient development of new knowledge, albeit only perhaps
of sub-optimal quality. The end of the latter was the perfection of knowledge—
specifically, the perfection of the imperfect knowledge produced by the justi-
ficatively sub-optimal methods of discovery.10
Traditionally, then, ‘discovery’ and ‘invention’ were synonyms. They
marked the first stage of a classical two-staged conception of justification
which was divided into a certificative stage and an argumentative stage. The
general idea behind this Classical Scheme can be summarized as follows:
Genuine scientific justification of a proposition11 requires both certifica-
tion and argumentation. Argumentation by itself is not enough because
it can too easily be fictional. De-fictionalization of argumentation is
the proper job of certification. Genuine demonstration, then, is certi-
fied argumentation, and genuine scientific knowledge requires genuine
demonstration.
This scheme figured significantly in the thinking of ancient geometers. An
interesting example is Proclus (411–485) who, in his Commentary on Book
I of the Elements, appealed to the Classical Scheme to explain the ‘ordering’
of Propositions I–IV.12
IV: Two triangles having two of their respective sides equal, and the angles con-
tained by those sides equal, will also have equal bases, be equal to each other as
triangles, and have the remaining angles equal.
13 Proposition IV is also about equal triangles, angles and equal angles. This being
so, one would naturally expect to see preliminary propositions establishing the exis-
tence of these items. Proclus did not comment on this and we can therefore only wonder
whether he regarded it as a defect in Euclid.
84 M E A N I N G I N M AT H E M AT I C S
In saying that Propositions I–III are ‘rightly preliminary’ to Proposition
IV, Proclus seems to have meant that they address certain broad sceptical
challenges to it and/or its proof. Proposition I, for example, responds to the
challenge raised by the question ‘Do we know whether a triangle can be
constructed at all?’. Propositions II and III, on the other hand, respond to the
challenge that perhaps no straight line (or other geometrical quantity for that
matter) is equal to another.14
Overall, then, Proclus seems to have taken the view that fabricated theo-
rematic investigation is uncertifiable investigation or investigation that lacks a
subject. To avoid it, the objects that are to form the subject-matter of a would-
be theorem should be priorly shown to exist. This, Proclus believed, was (best?
only?) done by giving a genetic or constructive definition of the subject—a
definition which ‘explains the genesis of a thing; that is, how the thing is made
or done: as is this definition of a circle, viz., that it is a figure described by the
motion of a right line about a fixed point’ (Hutton, 1795–1796: 2000, Vol. 1,
p. 362).15
Modern views
Mathematicians and philosophers of the modern era held similar views, as the
following statement by Leibniz illustrates:
. . . the concept of circle put forward by Euclid—namely, that it is the figure
described by the motion of a straight line in a plane about one fixed end—
affords a real definition, for it is clear that such a figure is possible. It is useful
. . . to have [such] definitions. . . beforehand. . . [For] we cannot safely devise
demonstrations (secure texere demonstrationes) about any concept, unless we
know that it is possible; for of what is impossible or involves a contradiction
(impossibilibus seu contradictionem involventibus), contradictories can also
be demonstrated. This is the a priori reason why real definition is required for
possibility.
Leibniz (1683:1973, pp. 12–13 (294), brackets added)16
14 More specifically, Propositions II and III provided two basic methods for produc-
ing equal lines: one when no line previously exists at a given place, the other when a
longer line exists at that place (see Proclus, 1970, p. 183).
15 Similar ideas are familiar from western jurisprudence, in particular, from its
general adoption of the principle of corpus delicti. This requires evidence not only
for justified conviction but for justified trial as well. The idea is that for a trial to be
justified there must be evidence both of the ‘existence’ of a crime and of the identity
of its perpetrator. In the case of murder, this has typically taken the form of evidence
of death as the result of an act, and evidence of the criminal agency of an identified
person as its means. Other types of crimes call for other types of evidence of course.
All legitimate trial, however, requires evidence of both the existence of a crime and the
identity of the criminal.
16 The number in parentheses is the page number of the Latin text in Leibniz (1978,
Vol. 7). For a similar statement see Leibniz (1989, XXIV).
D I S C O V E R Y, INVENTION AND REALISM 85
Leibniz thus held that genetic or constructive definitions are valuable
because they make known a priori the possibility of the concepts they define
(see Leibniz, 1764: 1916, Bk. II, ch. II, §18). This in turn, he suggested, makes
demonstration ‘safe’ in a certain way.
He made a confusing variety of claims concerning impossible concepts and
their existence, however. Sometimes he suggested that there are impossible
concepts (i.e., concepts that imply a contradiction).
A concept is either suitable or unsuitable. A suitable concept is one that is
established to be possible, or not to imply a contradiction.
Leibniz (1903, p. 513)
Other times he suggested that impossible concepts cannot exist. Thus, for
example, he claimed that ‘we cannot have any idea of the impossible’ (Leibniz,
1978, Vol. 4, p. 424), and he also maintained that ‘what actually exists cannot
fail to be possible’ (Leibniz, 1978, Vol. 7, p. 214).17
My main concern here, though, is with the suggestion that there is some-
thing unsafe about demonstration concerning A when it is not accompanied by
knowledge of the possibility of A. Specifically, I’m concerned with the reasons
Leibniz may have had for believing this.
The danger he mentioned was a possibility of contradiction—‘of what is
impossible . . . contradictories can be demonstrated’ (loc. cit.). Taken as we
would take it today, though, this does not seem right. The principal type of
theorem in Leibniz’ time was a proposition of the form ‘All A are B’. The
contradictory of this would be (a sentence equivalent to) a sentence of the form
‘Some A are not B’, which requires the existence of A’s.
When A is impossible, though, A’s cannot exist. Hence, one who, like
Leibniz, maintained that the actual must be possible, would not have believed
it possible to demonstrate ‘Some A are not B’ for impossible A. Consequently,
he would have had no reason to regard demonstration of ‘Some A are not B’
as posing a threat to demonstration of ‘All A are B’ when A is impossible.
More likely, Leibniz was thinking of demonstration of ‘No A are B’
rather than demonstration of ‘Some A are not B’ as the imagined counter-
demonstration to demonstration of ‘All A are B’. Supposing this to be right,
though, we are left with the problem of saying what it is about such a situation
that would make it ‘unsafe’. Whatever else Leibniz may have thought the
danger to be, it would not be one of directly demonstrating a false statement.
That this is so follows from elementary (classical) logical facts. When A is
impossible, there are necessarily no A’s, and when this is so ‘All A are B’ and
‘No A are B’ are both true.18
17 See Mates (1986, pp. 66–68) for a brief but useful discussion of these puzzling
matters.
18 This supposes of course that A may still be a genuine concept even when it is
impossible. There were, of course, many who challenged this, on which more later.
86 M E A N I N G I N M AT H E M AT I C S
But if it is not the falsehood against which real definition of A protects
us, from what does it protect us? Leibniz was not explicit on this point.19 He
may have seen the threat as a threat of fictionalization—that is, the threat of
engaging in investigation that has no subject and hence, in the final analysis,
is about nothing and produces knowledge of nothing. This seems to have been
the classical view and was essentially the view Proclus was urging. Or so I
would argue. Others of Leibniz’ time, and of earlier and later times too, were
more definite, and they generally gave one of two views of the security issue,
one broadly practical, the other broadly theoretical in character. I’ll presently
describe and discuss each of these. For the moment, though, the main points to
bear in mind are that (i) there was and perhaps still is a tradition in the history
of mathematics of linking the reliable with the broadly experiential, at least to
the extent that construction was taken to be a medium of experience, but that
(ii) the content of such experience was not necessarily that of the proposition
ultimately being proved or justified.
Real Definition . . . explain[s] the genesis of a thing; that is, how the thing is
made or done: as is this definition of a circle, viz., that it is a figure described
by the motion of a right line about a fixed point.
Hutton (1795–1796: 2000, Vol. 1)
Frege too pressed the point in his disagreement with the creativists (e.g.,
Hankel and Hilbert). He thus responded to Hankel’s identification of consis-
tency and existence by observing;
Strictly, of course, we can only establish that a concept is free from contradic-
tion by first producing something that falls under it. The converse inference
is a fallacy, and one into which Hankel falls.
Frege (1884:1968, §95)
The question of the proper roles of nominal and real definition were thus
at the heart of the traditional discussion of the discovery vs. creation issue. It
remained a central element well into the twentieth century. Understanding this,
I believe, should help us both to deepen and to broaden our own perspective on
the issue. More particularly, it should lead us to an appreciation of what more
may lie behind the issue, and what more may be at stake in its resolution than
arguments like Gödel’s suggest.
Overall then, the theoretical defence of real definition in geometry was based
on two ideas. The first was that geometrical concepts are derived from obser-
vation of external objects by a process(es) of abstraction. The second was that
the contents of such observation are in an important sense given rather than
created. Real concepts are thus representations that begin with a content that
is given, not one that is created or fabricated. They then arise by a process of
90 M E A N I N G I N M AT H E M AT I C S
abstraction from this experience, a process which only subtracts from and does
not add to the original experienced contents.
This view of real concepts was sometimes extended to a general view
of concepts by nineteenth-century philosophers. A striking example was
Schopenhauer, who offered the following memorable account of the nature
of concepts and the conditions under which they exist.
. . . concepts derive their content from the intuitive realm and therefore the
entire structure of the world of thought rests upon intuitions. We must there-
fore be able to go back from every concept, even if indirectly through inter-
mediate concepts, to the intuitions from which it is itself abstracted . . . That
is to say, we must be able to support it with intuitions which stand to the
abstractions in the relation of examples. . .
These intuitions. . . afford the real content (realen Gehalt) of all our thought,
and whenever they are wanting we have not had concepts but mere words in
our heads (blosse Worte im Kopfe). In this respect our intellect is like a bank
which holds notes (Zettelbank), which, if it is to be sound, must have cash in
its safe, so as to be able to meet all the notes it has issued, in case of demand.
The intuitions are the cash and the concepts the notes.
Schopenhaver (1859: 1966, Book 1, ch. 7, vol. 2)22
Views of this general type were not, however, confined to philosophers or to
the nineteenth century. Indeed, no lesser nor less recent a mathematician than
Hermann Weyl used the image of notes of deposit to describe the difference
between genuine concepts and propositions, on the one hand, and purely
symbolic devices, on the other. He used the existence claim in mathematics
as his chief example, and he likened it to paper money—that is, money which
in itself has no value and whose only true value is that of a real commodity that
backs it.
An existential proposition (Existentialsatz)—something like ‘there is an even
number’—is not at all a judgment in the genuine sense of an assertion of
a fact. Existential states of affairs (Existential-Sachverhalte) are an empty
invention (Erfindung) of the logician. ‘2 is an even number’: that is a real
(wirkliches), a factual expression of a given judgment. ‘There is an even
number’ is only a judgment-abstract (Urteilsabstrakt) obtained from this
judgment. Just as I take knowledge (Erkenntnis) to be a valuable (wertvollen)
holding (Schatz), so I regard the judgment-abstract as paper (Papier), which
somehow indicates (anzeigt) a holding without disclosing (verraten) where
it is. Its only value can lie in its ability to get me to search for the hold-
ing. The paper is worthless so long as it is not realized (realisiert) by
a real (wirkliches) judgment such as ‘2 is an even number’ that stands
behind it.
Weyl (1921, p. 54)
22 See Schopenhauer (1911, p. 76) for German version. The first edition was pub-
lished in 1819, the second in 1844, a third in 1859. The passage quoted here is in all
these editions.
D I S C O V E R Y, INVENTION AND REALISM 91
In all the above we see the same basic idea at work, namely, that those
representations we call concepts exist only to the extent that they stand in a
relationship of abstractability to the contents of experiences that are given and
not created.
This offers a relatively clear explanation of how it is that proof of ‘No A are
B’ would trivialize or devalue proof of ‘All A are B’. If both ‘All A are B’ and
‘No A are B’ expressed propositional contents and were true, there would be
no A’s. Hence there would be no experiential content—that is, no experienced
instance of A—from which A could be abstracted. The expression ‘A’ would
therefore have no content and would not express a concept. As a result, ‘All
A are B’ would not express a proposition and could not therefore be a content
of genuine judgement. As a result, a demonstration of ‘All A are B’ would
not yield knowledge that all A are B, and, so, would not achieve its intended
purpose.
But though this reasoning is clear, it is also unconvincing, and for at least
two reasons. The first is its foreclosure of the possibility of uninstantiated
concepts. This is troubling since there seem to be genuine concepts (or at
least concept-like representations) that not only are not instantiated but, indeed,
are not capable of being instantiated. This at any rate for composite concepts
(e.g., that of a map requiring five colors for its colouring). Whether there are
concepts that can properly be regarded as primitive and are not instantiable is a
more difficult question to answer. I know of no good argument that there could
not be such contents, however.
The second reason why the Schopenhauer–Weyl argument is unconvincing
is its seeming blindness to the possibility that the type of experience capable
of sustaining derivation of a concept by abstraction might itself presuppose
the availability of concepts. The Schopenhauer–Weyl argument supposes that
we could have the types of contents needed to begin a process of abstraction
without having access to concepts that are not based on such acquisition. To
be convincing, though, a closer analysis of contents and abstraction would
have to be offered, and, with it, an argument that there are at least some types
of rudimentary contents we can grasp without prior application of concepts.
In addition, reasons would have to be given for the claim that there is a
plausible path of acquisition from such contents to mathematical contents via
abstraction.
Until such gaps are filled, the theoretical defence cannot readily be
accepted as a general account of the existence or acquirability of concepts.
But though it has shortcomings as a general theory of mathematical concepts,
the theoretical defence might yet provide a basis for a worthwhile distinction
between discovery and creation in mathematics. The idea would be that a real
concept is an instantiable concept. The existence of non-instantiable concepts
would not be denied. Neither would their use be forbidden. Their use, however,
would be controlled by an appropriate form of consistency requirement—
specifically, a consistency requirement the demonstrated satisfaction of which
would not necessitate instantiation of the concept(s) involved.
92 M E A N I N G I N M AT H E M AT I C S
Discussion
Both of the traditional defences of real definition discussed above offer possi-
bilities of significant distinctions between discovery and creation in mathemat-
ics. Both also diverge from Gödel’s phenomenological argument in that they
see the distinction between discovery and creation as wanted for more than
merely a phenomenologically adequate accounting of whatever experience we
may have of truth being ‘forced’ on us. It is not necessary that they deny that
there is a point behind Gödel’s phenomenological reasoning. Neither, though,
do they see it as indicating the primary role for experience in the development
of mathematical knowledge.
Both defences take the view that producing an instance of a concept is
fundamentally a matter of discovery rather than of invention. Moreover, they
both see discovery as discovery of reality where reality guarantees consistency
or possibility. Creation, on the other hand, does not. In order for real definition
to fulfill its intended justificative purpose, then, it is necessary that it be seen as
a case of discovery rather than creation. Otherwise, the mathematician would
possibly be faced with an endless regress of proper justificative demands, and
this is surely something she should want to avoid.
I mentioned above that the practical defence of real definition seems to
underestimate the variety of forms a consistency constraint on the use of cre-
ated concepts might take. This was due to the common belief prior to Hilbert’s
development of proof theory that the only way to prove the consistency of a
concept was to find a recognizable instance. Hilbert’s proof-theory provided a
syntactical alternative to this.
On the other hand, Gödel’s incompleteness theorems (particularly his sec-
ond theorem) suggest that there are limits to the applicability of the Hilbertian
alternative. This may mean that, practically speaking, we will often have to rely
on production of instances or construction of models to prove consistency. In
the end, then, Frege may have been right concerning the practical possibilities
of proving consistency.23 Therefore, in the final analysis, it is difficult to
determine how extensive a role for discovery the practical defence is capable
of supporting.
Conclusion
I would like to close by considering one final question concerning Gödel’s
proposed use of the phenomenon of given-ness or forcedness in mathematics.
This time, though, my concern is not whether we can reliably detect it and, if
so, with what degrees of confidence and discernment. Rather, it is with what a
proper response to experiences of ‘given-ness’ or ‘forcedness’ might or ought
to be in the first place, supposing that we do in fact have them. I was led to this
24 Dedekind raised similar concerns in arithmetic. In the first paragraph of Was sind
und was sollen die Zahlen? he thus warned against following the leadings of inner
intuition (innere Anschauung), and said that evident truths are quite often those most in
need of proof. This indeed became the leitmotif of this foundational work in arithmetic,
the central principle of which was that nothing at all capable of proof should be accepted
without proof, regardless of the degree of its evidentness. See Dedekind (1888), preface
to the first edition.
25 In this sentence, the reality of which Bolzano speaks is empirical or sensible
reality. He thus demotes the type of real definition found in Euclid from scientific
mathematics to what was then regarded as applied mathematics.
94 M E A N I N G I N M AT H E M AT I C S
same time, he likened forcedness in mathematics to sensory perception. This
would lead us to think that the axioms of set theory are to the larger domain of
mathematical truth what sense perceptions are to the larger domain of natural
scientific truth.
If this is correct, though, then, pursuing the analogy between mathematics
and natural science, the propositions of set theory that force themselves on
us as being true are not likely to be the basic laws of mathematics. That is,
they should not properly be taken to be axioms. Rather, they should be seen
as phenomena whose proper explanation requires deeper and more basic laws,
a theoretical mathematics if you will.26 It is not easy, however, to think what
these deeper laws might be. In addition, the analogy between the axioms of set
theory and sensory judgements does not square well with what Gödel himself
said about the relationship of set theory to number theory—namely, that the
former can be at least partially justified by its ability to simplify proofs of
theorems of the latter.
Overall, then, there are serious problems concerning the analogy between
sensory experience and the type of experience by which Gödel supposed
various truths of mathematics to force themselves on us as true. If forcedness
is symptomatic of experience, then the mathematical propositions forced on
us as true ought to be seen as the experiential part of mathematics. Pursu-
ing Gödel’s analogy we would then be led to consider the possibility of an
observation/theory divide in mathematics, similar to that which we accept
in natural science. Bolzano believed in something like this. Specifically, he
believed in objective differences of basicness between mathematical proposi-
tions. Moreover, he took it as the principal duty of the mathematician to reveal
this ordering of relative basicness.
Bolzano thus raised important questions concerning given-ness and its
possible significance to mathematical epistemology. I’ve indicated some of
the challenges these questions raise for Gödel’s attempt to introduce a type
of experience into mathematical epistemology. These should warn us against
equating degree of given-ness or forcedness with degree of basicness. They
do not, so far as I can see, speak with similar force against the traditional
understanding of the discovery/creation issue in mathematics—namely, that
which centers on the use of so-called ‘real’ definition in mathematics and the
protection it may provide against certain of the grosser forms of subjectivity.
I wish to explore, in a tentative and general way, the extent to which mathe-
matics is objective. As is typical in philosophy, part of our question is to try
to get clear on the meaning of the terms in the question. I take it for granted
that we know what mathematics is, or at least that we know it when we see
it, borderline cases aside. But what of ‘objectivity’? Intuitively, to be objective
is to be independent of human judgements, conventions, forms of life and the
like. But what of this notion of ‘independence’?
The view that mathematics is not objective—that mathematical truths are
somehow tied to the nature of human cognition, conventions or whatever—is
not uncommon among philosophers. Immanuel Kant took mathematics to flow
from ‘pure intuition’, the form of our faculty of perceiving the world spatially
and temporally. Thus, mathematics is directly connected to human abilities,
and so lacks objectivity—at least in some sense of that term. The traditional
intuitionists, L. E. J. Brouwer and Arend Heyting, follow suit. Heyting (1931:
1983, p. 52) wrote:
The intuitionist mathematician proposes to do mathematics as a natural func-
tion of his intellect, as a free, vital activity of thought. For him, mathematics
is a production of the human mind . . . [W]e do not attribute an existence
independent of our thought, i.e., a transcendental existence, to . . . mathemat-
ical objects . . . [M]athematical objects are by their very nature dependent on
human thought.
Today, we are in an even better position to appreciate the deep truth underlying
this than they were in the early decades of the seventeenth century. There is
hardly a branch of natural or social science that does not have substantial
mathematical prerequisites. One cannot get beyond the first few pages of a
textbook without considerable mathematical facility. Of course, this is applied
mathematics, while Horgan comments on pure mathematics. But the difference
between the more theoretical parts of some of the sciences and pure mathemat-
ics is not all that clear, at least to me.
Views, like Horgan’s, that deny the objectivity of even pure mathematics
make a mystery of Galileo’s observation—in a pejorative sense of ‘mystery’. If
mathematics is governed by no more than human conventions of assertibility,
as Horgan suggests, then why is mathematics so important for science? There
must be something about the WORLD, as it is in itself, independent of human
concerns, judgements, etc., that puts mathematics at the centre of just about all
of our attempts to understand it—even if they are our attempts to understand
the universe.
Galileo speaks of the language of the universe. Theology aside, I presume
that this is a metaphor. More literally, the claim is that one must invoke
mathematics in order to fully or properly understand the universe, at least in a
scientific manner. Does it follow that mathematical assertions themselves are
objective? It is probably also true that one cannot understand the universe at
any sophisticated depth without understanding a language. Does it follow that
M AT H E M AT I C S AND OBJECTIVITY 99
languages themselves are directly tied to the nature of the world, objectively,
independent of human interests, concerns, judgements, and the like? In any
case, there must be something about the non-linguistic world that makes
languages what they are, and make them effective in communication for us
humans. Similarly, there must be something about the non-mathematical world
that makes mathematics effective—indeed essential—for us to understand just
about anything.
To be fair, Heyting and Horgan are surely correct that mathematics and,
for that matter, language and science, are human activities, and the pursuit and
results of those activities are shaped by human concerns and interests. It is a
truism that theories and explanations, in both mathematics and science, are due
to both the nature of the non-human world and the nature of human knowers
and understanders. As John Burgess and Gideon Rosen (1997, p. 240) put it,
‘our theories of life and matter and number are to a significant degree shaped by
our character, and in particular by our history and our society and our culture.’
Of course, this is not to say that the world itself or, as Horgan or Putnam
might put it, THE WORLD ITSELF, is somehow shaped by ‘our character’. The
Burgess–Rosen observation is that it is our theories of the world that are so
shaped. The question before us is the extent to which the truths of mathematics
are due to the way the non-human world is, and the extent to which these truths
are due to the way the human mathematicians are.
Several competing philosophical traditions have it that there is no way to
sharply separate the ‘human’ and the ‘world’ contributions to our theorizing.
As Protagoras (supposedly) said, ‘man is the measure of all things’. On some
versions of idealism, not to mention some postmodern views, the world itself
has a human character. The world is shaped by our judgements, observations,
etc. And so, it would seem, there just is no WORLD, in Horgan’s and Putnam’s
sense. A less extreme position is Kant’s doctrine that the ding an sich, (or
DING AN SICH) is inaccessible to human inquiry. We approach the world
through our own categories, concepts, and intuitions. We cannot get beyond
those, to the world (or THE WORLD) as it is, independently of said categories,
concepts and intuitions.
On the contemporary scene, a widely held view, championed by W. V. O.
Quine, Hilary Putnam, Donald Davidson, and Burgess, has it that, to use a
crude phrase, there simply is no God’s eye view to be had, no perspective from
which we can compare our theories of the world to the WORLD itself, to figure
out which are the ‘human’ parts of our successful theories and which are the
WORLD parts.
This Kant–Quine orientation may suggest that there simply is no objectiv-
ity to be had, or at least no objectivity that we can detect. If this is right, then
there simply is no answering the question of this paper. For what it is worth,
I would resist this, despite sympathy with the Kant–Quine orientation. There
may not be such a thing as complete objectivity—whatever that would be—
but it still seems that there is an interesting and important notion of objectivity
to be clarified and deployed. There seems to be an important difference—a
100 M E A N I N G I N M AT H E M AT I C S
difference in kind—between statements like ‘pure water contains two hydro-
gen atoms for each oxygen atom’ and statements like ‘Broccoli is disgusting’
and ‘Manchester United is evil’. Our question concerns which side of this
divide contains mathematics.
Crispin Wright’s Truth and Objectivity (1992) contains an account of
objectivity that is more comprehensive than any other that I know of, providing
a wealth of detailed insight into the underlying concepts. Wright does not
approach the matter through metaphysical inquiry into the fabric of REALITY,
wondering whether THE WORLD contains things like moral properties or num-
bers. He focuses instead on the nature of various discourses, and the role that
these play in our overall intellectual and social lives. That is, Wright tries to
clarify what it is for us—the community of language users—to treat a stretch
of discourse as objective, as we attempt to negotiate and understand this world
we find ourselves occupying.
As Wright sees things, objectivity is not a univocal notion. There are
different notions or axes of objectivity, and a given chunk of discourse can
exhibit some of these and not others. The axes are labelled ‘epistemic con-
straint’, ‘cognitive command’, ‘the Euthyphro contrast’ and ‘the width of
cosmological role’. In a previous paper, Shapiro (2007), I argue that with the
possible exception of some troubling matters near the foundation, mathematics
easily passes all four tests. Mathematics is epistemically unconstrained: there
are unknowable truths. The Galileo observation attests to the extremely wide
cosmological role of mathematics: it figures in all sorts of explanations, most
of which are explanations of non-mathematical matters. Mathematics falls on
the Socrates side of the Euthyphro contrast—it is not response-dependent—
and mathematics easily satisfies cognitive command. In short, mathematics is
objective, if anything is.
On the other hand, the possible exceptions—the foundational matters—
loom large, since they go the heart of the Kant–Quine matter noted just above.
Here I want to revisit one of the axes, cognitive command, since it bears out
this theme, even more than I anticipated in the other paper.
According to Wright, cognitive command figures as an axis of objectivity
only if the discourse is epistemically constrained. So I will pause for a brief
sketch of that primary axis of objectivity.
Epistemic constraint is an articulation of Michael Dummett’s notion of
anti-realism. According to one of Wright’s articulations of this axis (1992,
p. 75), a discourse is epistemically constrained if, for each sentence P in the
discourse,
P ↔ P may be known.
In other words, a discourse exhibits epistemic constraint if it contains no
unknowable truths.
It seems to follow from the very meaning of the word ‘objective’ that
if epistemic constraint fails for a given area of discourse—if there are
propositions in that area whose truth cannot become known—then that dis-
course can only have a realist, objective interpretation:
M AT H E M AT I C S AND OBJECTIVITY 101
To conceive that our understanding of statements in a certain discourse is
fixed . . . by assigning them conditions of potentially evidence-transcendent
truth is to grant that, if the world co-operates, the truth or falsity of any such
statement may be settled beyond our ken. So . . . we are forced to recognise
a distinction between the kind of state of affairs which makes such a state-
ment acceptable, in light of whatever standards inform our practice of the
discourse to which it belongs, and what makes it actually true. The truth of
such a statement is bestowed on it independently of any standard we do or can
apply . . . Realism in Dummett’s sense is thus one way of laying the essential
groundwork for the idea that our thought aspires to reflect a reality whose
character is entirely independent of us and our cognitive operations.
Wright (1992, p. 4)
During the second half of the nineteenth century, through a process still
awaiting explanation, the community of geometers reached the conclusion
that all geometries were here to stay . . . [T]his had all the appearance of
being the first time that a community of scientists had agreed to accept in
a not-merely-provisory way all the members of a set of mutually inconsis-
tent theories about a certain domain . . . It was now up to philosophers . . . to
make epistemological sense of the mathematicians’ attitude toward geome-
try . . . The challenge was a difficult test for philosophers, a test which (sad to
say) they all failed . . . .
I take it that on the present scene, if Pat and Chris differ only over premises or
axioms, then they do not disagree at all. In effect, their words mean different
things. The two of them simply work in different structures. This explains why
mathematical theories are not discarded as false when they become unusable in
science, at least not now. Michael Resnik (1997, p. 131) calls the phenomenon
‘Euclidean rescue’.
Things may not be this neat if the disagreement concerns a more foun-
dational matter. Suppose that the ultimate conclusion of the ‘disputed’ proof
is a proposition of real analysis, but that Pat’s proof invokes a set-theoretic
principle, such as the continuum hypothesis or a large cardinal hypothesis, and
Chris rejects that set-theoretic principle. This would naturally focus the dispute
on the background set theory. One can invoke a Euclidean rescue there as
well, and say that Pat and Chris work in different structures, just because their
background set theories differ. Pat works in analysis-plus-set-theory-A, while
Chris prefers analysis-plus-set-theory-B. This sort of resolution is not quite as
comfortable as it was with, say, Euclidean and non-Euclidean geometries, due
to the pervasiveness of set-theoretic notions throughout mathematics, and the
M AT H E M AT I C S AND OBJECTIVITY 105
foundational role of set theory (see Maddy, 2007, pp. 358–360). The matter can
be murky, and it goes beyond the scope of this paper. We’d have to consider
whether mathematics can have more than one foundation, and, if it does, how
we’d study the relationships between the foundations.
In any case, a dispute concerning the set-theoretic background is much
like our one remaining possibility, that the difference between our mathemati-
cians’ attitudes traces to their logic. Suppose that Chris demurs from a rule of
inference that, for Pat, is so primitive that it cannot be broken down further.
To focus on an example, let us suppose that Pat’s proof invokes instances
of excluded middle, and Chris rejects that, since he is an intuitionist. This
raises the question of the objectivity of logic, which could (and did) take up
another lengthy paper (Shapiro, 2000). The details go beyond present concern,
but I suggest that with some qualifications, logic, too, passes the letter of
all of Wright’s tests for objectivity, with the possible exception of epistemic
constraint. But it is not clear what one should conclude from that.
One potentially troubling matter here there is that most of Wright’s axes of
objectivity are formulated in terms of logic, and so it seems that one can deploy
the various axes only after one has settled on a logic. That is, the various axes
of objectivity presuppose a logic (although it is left open just which logic is
presupposed). So it hard to see the extent to which we can even ask if logic is
objective, using Wright’s framework.
One might invoke Euclidean rescue with logic, taking an eclectic attitude
toward it. The thesis would be that, say, classical analysis and intuitionistic
analysis are two different subjects, and are no more in conflict with each other
than Euclidean and non-Euclidean geometries. If we go down this route, then in
a sense all of mathematics—once suitably idealized—reduces to calculation. It
is just a matter of what conclusions can be drawn in various deductive systems.
There is no content to mathematics beyond that. We save cognitive command,
but at the cost of there not being any interesting, genuine disputes among
mathematicians. The supposedly ‘disputing’ parties do not speak the same
language. We seem to have manoeuvered ourselves into some sort of formalism
concerning mathematics—at least once it is suitably idealized so that we can
apply the axes of objectivity. Of course, we are still putting Wittgensteinian
issues of rule-following aside.
Let us briefly explore the alternative position, that the classical mathemati-
cian (Chris) and the intuitionist (Pat) do have a genuine disagreement with
each other. And then we can ask if the matter is objective, along the axis of
cognitive command. More murky waters lie here.
Our question concerns whether one of our idealized disputants, Chris or
Pat, exhibits a cognitive shortcoming. ‘Inferential error’ appears in the list of
shortcomings that Wright gives in the above characterization of cognitive com-
mand.3 Surely Pat accuses Chris of inferential error. Chris invokes excluded
One starts with one’s own intuitions concerning logical correctness (or log-
ical necessity). These usually take the form of a set of test cases: arguments
that one accepts or rejects, statements that one takes to be logically neces-
sary, inconsistent, or equivalent . . . One then tries to build a logical theory
whose pronouncements accord with one’s initial considered judgements. It
is unlikely that initial attempts will produce an exact fit between the theory
and the “data” . . . Sometimes . . . one will yield one’s logical intuitions to
powerful or elegant systematic considerations. In short, “theory” will lead
one to reject the “data”. Moreover, in deciding what must give, not only
should one consider the merits of the logical theory per se . . . but one should
also consider how the theory and one’s intuitions cohere with one’s other
beliefs and commitments, including philosophical ones. When the theory
rejects no example one is determined to preserve and countenances none one
is determined to reject, then the theory and its terminal set of considered
judgements are in . . . wide reflective equilibrium.
Resnik (1997, p. 159)
“divergent input”, that is, the disputants working on the basis of different information
(and hence guilty of ignorance or error. . . ), or “unsuitable conditions” (resulting in
inattention or distraction and so in inferential error, or oversight of data, and so on), or
“malfunction” (for example, prejudicial assessment of data . . . or dogma, or failings in
other categories . . . ’ (Wright, 1992, p. 92)
M AT H E M AT I C S AND OBJECTIVITY 107
It seems not, although it is hard to see how one might construct an argument
for this. It seems to be possible that our two idealized mathematicians, Pat and
Chris, are both in reflective equilibrium, each by the lights of his or her own
logic. If so, how can we neutral observers accuse one of them of cognitive
shortcoming?
There is something troubling about the whole issue concerning the objec-
tivity of logic. Any serious dispute in any area of discourse is going to involve
logic. All disputants, in all areas, are themselves reasoners, and come to their
respective conclusions in part by drawing inferences. Given how pervasive
logic is, disagreements or differences about logic are certain to result in dis-
agreements or differences everywhere. If logic fails to be objective, can there
be any objectivity anywhere?
I apologize for failing to come to a crisp conclusion concerning the objec-
tivity of logic and, to that extent, the objectivity of mathematics. Given the
central role of logic in our theorizing, it is hard to separate it out for sharp
treatment. Any attempt to characterize how the question of objectivity is to be
adjudicated will presuppose logic.
The situation can be made more palatable if we recall the Kant–
Quine thesis, the idea that there is no way to sharply separate the parts of our
best theories that are due to the way the world is and the parts that are due
to the way that we, the human cognitive agents, are. One fallout, I submit, is
that there is no sense to asking for complete objectivity. So some feature that
compromises the objectivity of a given area of discourse does not eliminate
objectivity altogether. This is the nature of the holistic beast.
Later in his book, Wright (1992, p. 144) adds some qualifications to the
formulation of cognitive command. It looks like the qualifications are meant
to deal with matters like holistic adjudication, in our struggle for reflective
equilibrium. Wright says that a discourse exerts cognitive command if and
only if
it is a priori that differences of opinion formulated within the discourse,
unless excusable as a result of vagueness in a disputed statement, or in the
standards of acceptability, or variation in personal evidence thresholds, so to
speak, will involve something which may properly be described as a cognitive
shortcoming.
That is, Wright holds that blameless disagreement that turns on vagueness,
standards of acceptability and the like, does not undermine cognitive com-
mand. Why are there these exceptions? What happened to the original motiva-
tion for the criterion of cognitive command, which did not mention vagueness,
evidence thresholds, or the like, one way or the other? Is this an instance of
what Imre Lakatos (1976) calls ‘monster-barring’? We find some parts of our
theory that do not seem to fit, and so we just exclude them.
Although Wright does not put it this way, I suggest that the excep-
tions listed in the nuanced version of cognitive command are in line
with the Kant–Quine theme. On Wright’s views of vagueness—and mine
108 M E A N I N G I N M AT H E M AT I C S
(Shapiro, 2007a)—vague terms are response- or judgement-dependent, at least
in their borderline regions. He writes that it is ‘tempting to say . . . that a
statement’s possessing (one kind of) vagueness just consists in the fact that,
under certain circumstances, cognitively lucid, fully informed and properly
functioning subjects may faultlessly differ about it.’ Yet, robustly objective
areas of inquiry, such as natural science, are conducted with vague terms.
That, alone, cannot undermine their objectivity, unless one takes an all-or-
nothing approach to that question. Similarly, it is plausible that ‘standards
of acceptability’ and especially ‘variation in personal evidence thresholds’
lie closer to the ‘human’ side and further from the ‘world’ side of the mix.
Surely, a conservative scientist, one who is more cautious in putting forward
claims, need not have any cognitive shortcoming with respect to a slightly more
speculative colleague, nor vice versa. So a disagreement that is traced to that
difference need not undermine cognitive command.
The same may go for the more foundational matters concerning mathe-
matics and its logic. Disagreements that turn on holistic considerations may
end up being adjudicated, in part, by matters of taste concerning, for example,
what a given theorist finds to be elegant or simple. One mathematician may
prefer the subtle distinctions and sharper bounds produced by constructive
mathematics while another might go for the unity and, in some way, simplicity
and tractable meta-theory of classical mathematics. That is, it just may be
that certain foundational matters are negotiated closer to the ‘human’ than the
‘world’ side of the web. Even if this is so, it does not follow that mathematics
is not objective, even a paradigm of objectivity, one of the standards by which
we measure other discourses.
Comment on Stewart Shapiro’s
‘Mathematics and objectivity’
Gideon Rosen
The problem
The closing sentence of Paul Benacerraf’s famous paper is a kōan: a bit of
seeming nonsense that points—or seems to point—to a deep truth. It is a
theorem of basic arithmetic that there are two prime numbers between 15 and
20. Anyone who accepts basic arithmetic must therefore agree that there are
two prime numbers between 15 and 20, and hence there are at least two num-
bers, and hence that there are numbers. And yet the idea that numbers are real
things—that the real world contains mathematical objects in the same sense in
which it contains guns and rabbits—can sound preposterous or confused. And
so we find philosophers straining to articulate a position of the following sort:
Of course there are numbers (and functions and sets and mathematical
objects of other sorts). That’s just mathematics, and we have no quarrel with
mathematics. But in another sense—the metaphysical sense—there are no
numbers. Numbers are not real. Numbers are not Things.1
It should be obvious that such remarks are seriously puzzling as they stand.
Suppose your exterminator tells you that you have squirrels in your attic,
but then goes on to add that in the strict and philosophical sense there are
no squirrels. Or suppose an astrophysicist reports that there are three black
2 I might simply have said that I use ‘exist’ and the other existential idioms to mean
what they mean in ordinary English. As Quine stressed, there is no good reason to
believe that these words mean one thing in mathematics and something else in other
areas (Quine, 1960, §27). Numbers are plainly very different from tables and quarks
and mental images. But if all of these things exist—i.e., if there are such things—there
is no reason to suppose that they exist in different senses.
3 Some philosophically minded mathematicians have expressed doubts about certain
parts of classical arithmetic—the impredicative or non-constructive parts (for example,
Nelson, 1986). These specialized disputes need not concern us. Constructive arithmetic
involves existence claims, e.g., the claim that there are two prime numbers between 15
and 20.
116 M E A N I N G I N M AT H E M AT I C S
than in the arithmetical claim that there are two prime numbers between 15
and 20. But if it is a matter of choosing between grand metaphysical schemes
and basic arithmetic, it is clear to me that it is the metaphysics that ought to
budge.
Other philosophers reject basic arithmetic because they think that if num-
bers and other mathematical objects existed, there would be no way for us
to know anything about them. This sort of claim is typically supported by
a general philosophical theory of knowledge, according to which knowledge
requires some sort of interaction between the would-be knower and the object
of his inquiry (Benacerraf, 1973). These theories were originally developed
to account for empirical knowledge, and they may be useful for that purpose.
But if the philosopher insists that they hold in complete generality, he faces a
glaring difficulty. These restrictive theories typically entail that the usual ways
of fixing opinion in mathematics—calculation, proof, informal mathematical
argument—could not possibly be sources of knowledge (since they do not
involve causal intercourse with the numbers), and hence that someone who has
come to believe in the ordinary way that 235 + 657 = 892 does not really know
that 235 + 657 = 892. But then the question arises: Why isn’t this simply a
counter-example to the philosopher’s theory? In other areas, when a philosoph-
ical theory is incompatible with an otherwise uncontroversial fact, the normal
response is to reconsider the theory or to restrict its scope. A philosophical
theory of knowledge must accommodate the manifest fact of mathematical
knowledge. If the theory clashes with this fact, so much the worse for the
theory.
This is just a rough sketch of an intricate dialectic, but the main tactic
should be clear. The central claims of mathematics are unassailable by all
pertinent mathematical, scientific and commonsensical standards. Any philo-
sophical challenge to these claims is thus a sceptical challenge of a certain
special sort, one that depends on bringing distinctively philosophical principles
to bear on questions that are not themselves purely philosophical. Sceptical
challenges of this sort are notoriously weak. When speculative philosophy
contradicts settled science or common sense, the normal response—and, I
believe, the reasonable response—is to suspect that it is the philosopher who
is mistaken. This is not a firm principle, but it is a good rule of thumb. And if
we follow it in this case we have no choice but to allow that since there are two
prime numbers between 15 and 20, mathematical objects therefore exist.
4 ‘Arithmetic as the natural history (mineralogy) of numbers. But who talks like this
about it? Our whole thinking is penetrated with this idea.’ (Wittgenstein, 1956: 1967,
IV, p. 11)
5 The view has few contemporary adherents among philosophers, though mathe-
maticians often find it congenial. See Curry (1951) for one version. For an exposition
and assessment of Frege’s celebrated objections to formalism, see Resnik (1980).
6 The axioms include the basic principles governing the successor function:
0 is a number
0 is not the successor of any number
Every number has a successor
No two numbers have the same successor;
the recursion equations for addition and multiplication;
For any number x, x + 0 = x
For any numbers x and y, x + the successor of y = the successor of (x + y)
118 M E A N I N G I N M AT H E M AT I C S
with an infinitary rule of inference—an omega rule—that permits the infer-
ence from an infinite list of premises A(0), A(1), . . . A(n) . . . to the universally
quantified conclusion: For all numbers x, A(x). PAω is obviously a sound
theory. The axioms are true, and the rules of inference preserve truth. More
importantly, PAω is also a complete theory, in the sense that every sentence A
in the language of arithmetic is such that either A or its negation is provable in
PAω .7 This means that anyone who accepts standard arithmetic should accept
the following equivalence:
For any sentence A in the language of arithmetic, A is true if and only if A is
provable in PAω .
Taken by itself, this equivalence is a mathematical fact with no special
metaphysical significance. Even the most unreconstructed Platonist should
accept it. But now consider the formalist’s characteristic claim:
For any true arithmetical sentence A:
A is true in virtue of the fact that A is provable in PAω ; or
What makes A true is the fact that it is provable in PAω ; or
A’s truth is grounded in the fact that A is provable in PAω ;
A’s truth consists in the fact that A is provable in PAω .
The italicized idioms here are not part of the official vocabulary of mathemat-
ics. Mathematics assures us that claims about the natural numbers are equiv-
alent to claims about the formal provability of certain sentences in a certain
formal system. But the formalist’s distinctive claim is that the arithmetical
facts are somehow grounded in these proof-theoretic facts, and that the proof-
theoretic facts are therefore, in a corresponding sense, more fundamental. From
the standpoint of mathematics, this claim is extracurricular. It is a distinctively
philosophical claim about the metaphysics of arithmetic.
Note that even though the reductive formalist regards certain linguistic
items (sentences and formal systems) as more fundamental than the numbers,
he cannot deny that numbers exist. Since the existence theorems of arithmetic
(e.g., there are two primes between 15 and 20) are all provable in PAω , the
would have been another instance of this pattern. We call any ordered collec-
tion of this sort an omega sequence. In general, an ordered collection is a pair
(X, ≺) consisting of a set X and a relation ≺ on that set. In the hypothetical
example, the Roman emperors together with the temporal relation x’s reign
precedes y’s reign constitutes an omega sequence, as do the natural numbers
together with the relation x is less than y.
In order to state the relevant fact about arithmetic, we need one more bit
of received wisdom. It is well known that given a suitable logical background,
every arithmetical claim that can be formulated in the elaborate technical lan-
guage of modern number theory can be expressed in a stripped-down language
in which the only primitive symbols are N (for natural number) and < (for
the standard less than relation). If A is an ordinary arithmetical sentence (e.g.,
there are two prime numbers between 15 and 20), let’s call its translation into
this stripped-down language A[N, <].
The pertinent theorem is then as follows:
For any claim A in the language of arithmetic, A[N, <] if and only if, for any
omega sequence (X, ≺), A[X, ≺].
A[N, <] if and only if as a matter of necessity, for any omega sequence
(X, ≺) A[X, ≺].
THE R E A L I T Y O F M AT H E M AT I C A L O B J E C T S 121
Even if there are no infinite sets in the actual world, the claim that any
possible omega sequence has a certain structural feature A(X, ≺) is non-trivial
(on the assumption that omega sequences are possible). The theorem affirms
that this modal claim is equivalent to the ordinary mathematical claim A
with which we began. (A modal claim is a claim about what is possible or
necessary.)
Now so far this is just a bit of uncontroversial (if somewhat unfamiliar)
mathematics. Anyone who accepts standard arithmetic should accept this
equivalence. The structuralist’s characteristic philosophical claim is that the
truths of arithmetic reduce to or are grounded in general claims about all
possible omega systems.10 Return to our example, the claim that there are two
prime numbers between 15 and 20. The structuralist accepts this claim, since
he accepts ordinary arithmetic as it stands. And this means that he (like the
formalist) cannot deny that numbers exist. His distinctive claim is that this fact
about the numbers is grounded in the fact that if there were an omega sequence
of whatever kind, it would have a certain complex structural property—very
roughly, the property of having two ‘prime elements’ in between in its 16th
and 21st elements.11 And here is the crucial point: on the face of it, this latter
fact is not a fact about a distinctive kind of object. It is not a fact specifically
about numbers, nor is it a fact specifically about Roman emperors. Indeed there
is a sense in which it is not a fact about anything at all. After all, a conditional
modal claim of the form if there were an infinite set of a certain kind, it would
exhibit such and such features does not affirm the existence of anything in the
actual world. According to one sort of structuralist, the truths of arithmetic
are made true by modal facts of just this sort. And it is not hard to see why
someone who takes this view might be tempted to say that there is a sense
in which the numbers are not real things. The truths about penguins are made
true by the birds and their behaviour. The truths about the numbers, by contrast,
are not really made true by the numbers; rather they are made true by general
conditional facts in which the numbers themselves make no appearance.
12 For my purposes, facts might be identified with true structured propositions of the
sort originally described by Bertrand Russell (1905).
THE R E A L I T Y O F M AT H E M AT I C A L O B J E C T S 123
obtains in virtue of another, or that one fact makes another fact obtain. Some
examples may help:
Disjunctive facts are grounded in their true disjuncts. It is a fact that I am
now either in New Jersey or in Cambridge. The fact obtains (as it happens)
in virtue of the fact that I am in New Jersey. If I had been in Cambridge, the
same fact would have obtained for a different reason.
Existential facts obtain in virtue of their instances. It is a fact that someone
spilled the milk. As it happens, this fact is grounded in the fact that Fred
spilled the milk. If someone else had spilled it, the same fact would have
obtained for a different reason.
Facts about the determinable features of things are grounded in more deter-
minate facts. A certain ball is red in virtue of being (say) crimson; the
particle has a mass of between 10 and 20 MeV in virtue of the fact that
its mass is exactly 17.656 MeV.
Facts involving definable properties and relations are grounded in their
‘definitional expansions’. To be a square is, by definition, to be an equilateral
rectangle (or so we may pretend). Given this, if ABCD is a square, then it is
square in virtue of the fact that it is equilateral and rectangular. That is what
makes the figure in question a square.
Supervenient facts are typically grounded in the facts upon which they
supervene, even if we cannot state the patterns of dependence in a systematic
way. The fact that the US trade deficit with China in 2008 was roughly $117
billion supervenes upon a vast mosaic of particular facts about individual
economic transactions, and perhaps ultimately on some vaster array of facts
about the quarks and electrons that composed the people who did the buying
and selling. Macroeconomic facts obtain in virtue of these lower-level facts,
even if it is impossible in practice, and perhaps even in principle, for us to
spell out the micro facts in virtue of which any given macro fact obtains.
These examples bring out two important features of the grounding relation.
First, it is a form of necessitation. The facts that ground a given fact entail
the fact they ground as a matter of absolute necessity. This distinguishes the
grounding relation from certain forms of causal or nomological determination.
There is no doubt a sense in which effects are grounded in their causes: the
cause makes the effect occur, and so on. But as Hume noted, it is always
possible for the cause to occur without the effect. The grounding relation
that interests us involves a much more intimate form of dependence, as the
examples show.
Second, the grounding relation is an explanatory relation. To cite the facts
that ground a given fact is to give information about why that fact obtains. To
suppose that the grounding relation is an objective relation, as I do, is therefore
to suppose that there are objective facts about the explanatory order (which is
not to say that the practice of giving explanations is always a matter simply of
reporting these objective facts).
124 M E A N I N G I N M AT H E M AT I C S
These remarks do not amount to a definition of the grounding relation. My
own view is that the relation is too basic to admit of definition. More might be
said by way of informal explication, of course, but I hope that this much will
suffice for our present purposes.13
The proposal
A reductionist proposal in the philosophy of mathematics is a claim to the
effect that every mathematical fact of a certain kind—e.g., every truth of
arithmetic, or every truth of set theory—is ultimately grounded in facts of a
rather different kind, e.g., acts about formal provability, or about what would
have been the case if there had been an infinite sequence of objects.14 We have
considered two examples in which a reductionist proposal of this sort seems to
imply that the objects of arithmetic are somehow less real or less ‘thing-like’
than certain other things. And in both of these proposals the reduction has had
a distinctive character: the objects of the higher-level theory do not figure in the
more fundamental facts to which that theory has been reduced. This suggests
a natural strategy for explaining the metaphysical thesis of qualified realism.
Let us say that a fact is fundamental if it is not grounded in further facts,
and that a thing is fundamental if it is a constituent of a fundamental fact.
Then we might identify full-strength realism about mathematics with the thesis
that some mathematical objects are fundamental things. This is certainly the
view of the hardcore Platonist for whom the numbers are sui generis abstract
substances—invisible luminous spheres arrayed in Plato’s heaven. But it is also
the view of more moderate Platonists who regard arithmetic, or perhaps some
more comprehensive theory like Zermelo-Frankel set theory, as an autonomous
body of truths not grounded in anything more basic. The forms of qualified
Qualified realism about F’s is the thesis that F’s exist, but no fundamental
fact contains an F as a constituent.
When a philosopher tells you that numbers are not real in the metaphysical
sense (even though they exist in the mathematical sense) one thing he or
she might mean is this: that every fact in which a number figures ultimately
obtains in virtue of some collection of facts in which numbers do not figure as
constituents.
The idea gains support from reflection on the ontological status of ‘emer-
gent’ or ‘higher-level’ entities in other areas. Consider the US dollar, or the
European Union or the information encoded on my hard drive. There is a
straightforward sense in which these things exist. The dollar exists as a genuine
form of currency in the sense in which the Italian lira once existed but no
longer does. There is a European Union in the sense in which there is not,
but might one day be, an Intergalactic Federation. My hard drive really does
encode a certain body of information—information that might be destroyed if
I press the wrong sequence of keys. And yet there is a powerful tendency to
think that while these existential claims are all perfectly correct, it would be
a mistake to think of the US dollar as a Thing in the sense in which a table
or God—or a number as conceived by the Platonist—would be a Thing. The
proposal allows us to make sense of this tendency. It is very natural to suppose
that every fact about the dollar or the EU is ultimately grounded in facts about
things of a different sort—facts about the attitudes and actions of economic
actors and the gold reserves in Fort Knox; facts about the legal arrangements
among EU states; etc. Of course there is little hope of stating in any finite
way the complete account of that in virtue of which the dollar is currently
valued at 0.6825 euros. But the proposal does not require that. Unlike some
earlier conceptions of reduction, reduction in our sense is not a thesis about the
meanings of sentences in the higher-level vocabulary, or about the possibility
of translating such sentences into another idiom. It is simply the claim that
every fact about currencies and the like is ultimately grounded in some perhaps
infinitely complex pattern of facts about other things. Now of course monetary
reductionism of this sort may not be true. That is a substantive question in the
philosophy of economics. The proposal is simply that our intuitive sense that
the US dollar is ontologically ‘derivative’ or ‘second-rate’ derives from our
strong suspicion that the dollar is unlikely to be a fundamental thing in the
sense defined above.
As formulated, the proposal entails that there are bona fide Things only if
there are fundamental facts. And it may be objected that we cannot assume this
126 M E A N I N G I N M AT H E M AT I C S
a priori. For all we know, there may be infinite descending chains or trees in
which P is grounded in Q and R, which are in turn grounded in S, T, U and V,
and so on ad infinitum.
Q R
S T U V
[A] obtains in virtue of. . . (something about the thoughts, activities or prac-
tices of conscious beings).
In fact this is a good general scheme for idealism and constructivism in
every area of philosophy. Berkeley’s idealism about external objects might be
understood as the claim that every fact involving tables and chairs is ultimately
grounded in facts about the mind of God. Mill’s secular idealism would substi-
tute facts about possible human sensations. Kantian Constructivism in ethics
128 M E A N I N G I N M AT H E M AT I C S
might be the thesis that every fact about right and wrong is ultimately grounded
in facts about the judgements to which any rational agent is committed when
he deliberates about what to do. Social constructivism of the sort that used
to be fashionable in the humanities might be the view that facts about social
reality (and in the absurd limiting case, facts about reality tout court) are
ultimately grounded in facts about our epistemic practices for accepting and
rejecting claims about that reality. These views are sketchy in the extreme,
but they all have the right general form: facts of one sort—facts which do not
seem to concern human thought directly—are said to be grounded in facts
about human thought or practice. The characteristic objects of the higher-
level discourse—ordinary objects, social classes, moral obligations, etc.—all
disappear when we examine the underlying facts, to be replaced by items of
a different sort: representations, ideas, people and their practices, etc. These
radical forms of idealism may be mistaken. But they exert a perennial (and
mysterious) attraction for philosophers and philosophically minded scholars
in other disciplines, so it is presumably worth seeking a clear statement of
them, as well as an account of why they strike us as incompatible with an
unqualified realism of the domain in question. The present framework provides
the beginnings of such an account.15
Let us return now to the mathematical case. Our framework puts us in a
position to entertain reductionist theses of the form:
For every mathematical fact [A] in some area:
[A] obtains in virtue of (. . . some fact that does not involve mathematical
objects as constituents).
But having entertained such theses, how are we to assess them? As our
examples begin to show, there are many ways to associate the arithmetical
facts with facts in a more fundamental-seeming idiom—proof-theoretic facts,
set-theoretic facts, facts of pure modal logic, etc.—in which the objects of
arithmetic do not figure as constituents. These proposals are all materially
adequate: they pair truths with truths; and when properly constructed, they
preserve logical relations. Indeed, in many cases they respect an even more
stringent (though somewhat elusive) constraint, which we might call relevance.
As noted, any proof of an arithmetical statement A is readily converted by
trivial steps into a proof of its formalist or modal structuralist counterpart,
and vice versa. Speaking strictly there is a difference between the fact that
there are two primes between 15 and 20 and the fact that every possible omega
16 For a survey of reductionist proposals for analysis, see Burgess and Rosen (1997,
ch. B).
130 M E A N I N G I N M AT H E M AT I C S
formalist numbers, facts about which are grounded in facts about provability in
PAω , the modal structuralist numbers, facts about which are grounded in facts
about all possible omega systems, and so on. Since the differences between
these systems of numbers make no mathematical difference, the language and
practice of mathematics will have had no occasion to distinguish them. If this
is right, then ordinary mathematical language will be rife with semantic inde-
terminacy. When I point to a river in the distance and say, ‘Let’s call that river
“Alph” ’, I introduce a meaningful word; but since I have not bothered to deter-
mine which of the many river-like objects in the vicinity I mean to pick out,
my word does not denote a single thing. Rather it ‘divides its reference’ over
a range of candidates—some a bit wider than others, some a bit longer, etc.
When I use the word in a sentence, I simply have not made up my mind about
which of these candidates I mean to denote. And in such cases it is misleading
to speak of the fact that Alph is in Xanadu. Since there are many river-like
objects in the vicinity, there are many such facts, the differences between which
simply do not matter for my purposes. Likewise, if there are many systems of
number-like objects distinguished only by the ways in which the facts about
them are grounded and not by any mathematically important features, it makes
no sense to speak of the fact that 235 + 657 = 892. There are rather many
equally qualified facts in the vicinity, each concerning numbers of some deter-
minate kind, each grounded in some determinate way in underlying facts.17
If this is the metaphysico-semantic situation, then it is no wonder that
we do not know how to answer questions about how the facts of arithmetic
are ultimately grounded. We do not know how to answer these questions
because they have a false presupposition, viz., that there is a unique system
of mathematical objects and determinate range of facts about them to which
the ordinary language of arithmetic manages to refer.
It is far from certain that this combination of metaphysical pluralism and
semantic indeterminacy is ultimately coherent. In order to address the issue
we would need a general theory of the conditions under which facts of one
kind give rise to or generate facts of another kind, and the project of producing
a general theory of this sort is genuinely daunting. But let us suppose for a
moment that the view is not only coherent but correct. We may then note that
even though questions about the numbers are to be rejected as badly posed, we
might still endorse a form of qualified realism about arithmetic. For it might
turn out that every candidate interpretation of the language of arithmetic takes
that language to describe a class of objects that disappear upon reduction. If
that is so, then it will still be true to say that arithmetical objects are not to be
Rosen says that he does not himself endorse this view, but it resonates with
me in the way that he wants it to: I think that if I were to try to put forward a
full defence of my own qualified realism about numbers, I would indeed do so
by trying to show that facts about numbers are grounded in other facts. There
are several ways that one might go about this, just as there are several different
philosophical positions that fall under the banner of qualified realism. Rosen
discusses two of them in detail (neither of them the approach that I myself
would follow), and also discusses the coherence of the position in general.
One of the merits of Rosen’s proposal is that it replaces murky questions
about the reality of numbers and mathematical statements (murky because it
is often not clear what ‘real’ means) by the closely related but much clearer
question of whether all mathematical facts are grounded in other facts. To
answer this question is a large project, but it is also a clearly defined one that
philosophers, and perhaps even mathematicians, could get their teeth into.
This page intentionally left blank
10
Getting more out of mathematics
than what we put in
Mark Steiner
In The Emperor’s New Mind (Penrose, 1989), and more recently, in The
Road to Reality (Penrose, 2005), Professor Penrose has championed what he
calls Platonism, Plato’s mathematical world. Since Professor Penrose is to be
present at this Symposium, I thought it would be appropriate to discuss his
view, or, rather what I think his view should be, based on the thrust of his
published work.
What Penrose is after, as he explains in Road, is the objectivity of
mathematics1 —rather than the existence of ‘mathematical objects,’—whether,
as another speaker at this Symposium, Professor Rosen, puts it, mathematics
is a ‘subject with no object’. As Burgess and Rosen point out in their excellent
book (Burgess and Rosen, 1997), Platonism was hijacked by Quine, who
defined it as ‘quantifying over mathematical objects’, and it no longer means
what it used to mean. Book after book (most of them published by Oxford
University Press) appear with learned discourses about, for example, whether,
should ‘mathematical objects’ cease suddenly to exist, it would make any
observable difference.2
I think, therefore, it would be better to eliminate reference to Plato alto-
gether, and speak of concepts, rather than objects, and look to Descartes as the
1 A modern source for the idea of emphasizing ‘objectivity’ over ‘objects’ in the
philosophy of mathematics are the writings of Georg Kreisel.
2 This question, by the way, is meaningless according to Quine’s philosophy, since
according to his form of Platonism we cannot describe even the observable world
without ‘quantifying over’ mathematical objects. The indispensability of atoms is
different—we can describe the appearances without appeal to atoms, but we need
the atoms to explain those appearances. Quinian indispensability is, as my late and
lamented teacher Sidney Morgenbesser used to say, a form of Kantianism, which Quine
probably inherited from C. I. Lewis, and thus it is misleading to speak of the ‘Quine–
Putnam indispensability thesis’; Putnam uses indispensability in the latter sense.
136 M E A N I N G I N M AT H E M AT I C S
real source of Penrose’s feeling that such concepts as that of the Mandelbrot
set are objective. After all, Penrose does not think that every concept is
objective (as did Plato), and neither did Descartes, as he puts the matter in
Meditation III:
. . . we must notice a point about ideas which do not contain true and
immutable natures but are merely ones which are invented and put together
by the intellect. Such ideas can always be split up by the same intellect, not
simply by an abstraction but by a clear and distinct intellectual operation,
so that any ideas which the intellect cannot split up in this way were clearly
not put together by the intellect. When, for example, I think of a winged horse
or an actually existing lion . . . I readily understand that I am also able to
think of a horse without wings, or a lion which does not exist . . . and so on;
hence these things do not have true and immutable natures. But if I think of a
triangle or a square . . . then whatever I apprehend as being contained in the
idea of a triangle—for example that its three angles are equal to two right
angles—I can with truth assert of the triangle . . .
What Descartes is saying here is that you get more out of mathematics
than what you put in; there is ‘latent information’ inherent in mathematical
ideas that is not contained in their verbal definition.3 This ‘latent information’
is part of the essence of these ideas, not put there by any mathematician. This
is, in my opinion, the real difference between mathematics and games. Take
the following position in a chess endgame (see Fig. 10.1).4
White needs to play 262 accurate moves to mate, assuming perfect play
on both sides. In many cases the moves ‘make no sense’, in the sense that one
cannot explain them without actually giving the entire tree of moves. Though
obviously this kind of thing is not ‘expected’ (the 50 rule move, which is
violated five times in the winning ‘line’, proves this), the surprise here is how
little is contained in the rules of chess, how little you reap for the investment.
Thus, the essence of chess is arbitrary, certainly not one of Descartes’ ‘true and
immutable essences’.
Compare this to one of Professor Penrose’s favorite examples, the ‘magi-
cal’ complex numbers. When they were introduced by the Italians as imaginary
solutions for equations over the reals, nobody could have predicted that they
would play the role of tying together the real numbers, as in the beautiful
equation
eπ i + 1 = 0
3 I think Descartes is saying more than just that some mathematical truths are
synthetic in the sense of Kant. For Kant, both 7 + 5 = 12 and the theorem about the
sum of the angles of a triangle are synthetic truths, but in the first case we don’t learn
anything of mathematical interest, since the sum had to be either one number or another.
The sum of the angles of a triangle is real information.
4 See http://www.chessbase.com/newsroom2.asp?id=239; thanks to Sylvain Cap-
pell for supplying this URL.
GETTING M O R E O U T O F M AT H E M AT I C S 137
A B C D E F G H
Note that when imaginaries were introduced, the idea of raising a real
number to an imaginary power was unthinkable even to Cardano and Bombelli.
Yet once one thinks of the idea, it turns out that there is little or no choice in
how to proceed.
Another property not recognized by the originators of imaginary numbers
is the absolute value of a complex number, which emerges when we embed the
complex numbers in the Euclidean plane. Using this property, mathematicians
can explain facts about the real numbers; for example, why the real function
1
1+x2
, defined everywhere on the reals, is not equal to its expansion as a power
series 1 − x2 + x4 − · · · wherever |x| ≥ 1 (the complex numbers with absolute
value 1 form a circle around the origin, and the real number 1 and the imaginary
i lie on this circle. For i, the function, even when continued into the complex
plane, is not defined, since the denominator is zero. Standard theorems in
complex analysis do the rest).
One might say that the introduction of imaginaries is motivated in part by
computational convenience: Cardano uses imaginary numbers even in comput-
ing real roots of a cubic equation (which can appear in his famous formula).
138 M E A N I N G I N M AT H E M AT I C S
Even so, the ‘latent information’ inherent in his idea goes far beyond calcula-
tional convenience. All in all, the following passage from The Emperor’s New
Mind could have been written by Descartes (had he been able to prophesy the
future of mathematics):
While at first it may seem that the introduction of such square roots of negative
numbers was just a device—a mathematical invention designed to achieve a
specific purpose—it later becomes clear that these objects are achieving far
more than that for which they were originally designed. As I mentioned above,
although the original purpose of introducing complex numbers was to enable
square roots to be taken with impunity, by introducing such numbers we find
that we get, as a bonus, the potentiality for taking any other kind of root
or for solving any algebraic equation whatever. Later we find many other
magical properties that these complex numbers possess, properties that we
had no inkling about at first. These properties are just there. They were not
put there by Cardano, nor by Bombelli, nor Wallis, nor Coates, nor Euler, nor
Wessel, nor Gauss, despite the undoubted farsightedness of these, and other,
great mathematicians; such magic was inherent in the very structure that
they gradually uncovered. When Cardano introduced his complex numbers,
he could have had no inkling of the many magical properties which were
to follow—properties which go under various names, such as the Cauchy
integral formula, the Riemann mapping theorem, the Lewy extension prop-
erty. These, and many other remarkable facts, are properties of the very
numbers, with no additional modifications whatever, that Cardano had first
encountered in about 1539.
Penrose (1989, pp. 96–97)
5 I am grateful to Sylvain Cappell and Stanley Ocken for improving the formulation
here. Professor Cappell brought to my attention a deep result of Frank Adams, the
English topologist, that only in dimension 1, 2, 4 and 8 does Euclidean space have a
vector space with a multiplication that satisfies: (a) there is a multiplicative identity
(left and right); (b) the product of two non-zero vectors is non-zero. For n = 8, the
multiplication is not associative.
140 M E A N I N G I N M AT H E M AT I C S
explicitly the homomorphism between SU(2), and the rotation group is 1884,
when Felix Klein (1888, p. 34ff)6 lectured on the icosahedron group and two
years later used the homomorphism to calculate with rotations in his study of
the gyroscope (see Klein, 1922).
SU(2) turns out to be isomorphic to the unit quaternions, so there is also
a two-to-one homomorphism from SU(2) onto all the rotations. For example,
the SU(2) matrices
iθ
e 0
0 e−iθ
correspond to a rotation of 2θ around
the z-axis7 When θ = 0, the matrix
10
is the identity matrix I = corresponding to the null rotation; when
01
θ reaches π , the matrix is −I, and the rotation is a full rotation. It takes
two full rotations to bring the matrix back to I. Again, we have a kind of
gratuitous labelling of the rotations as positive or negative, rotations in the
interval (360◦ , 720◦ ) being with SU(2), and calculational convenience was the
rationale of introducing SU(2) in the first place.
What I find remarkable is that this very superfluous information turns out
to be the key to some of the fundamental features of our universe. For the
symmetry of the electron turned out precisely to be that of SU(2). When we
rotate the electron 360◦ its quantum mechanical description (wave function) is
multiplied by −1! To get the electron back to its initial state, one must give it
two full turns. Had mathematical physicists been much better calculators, they
might have missed one of the spectacular discoveries in the history of science.8
This is one of many examples in which human limitations, far from impeding
scientific progress, were responsible for it.9
6 Klein does not explictly state that spatial rotations can be represented by 3 × 3
real orthogonal matrices with determinant +1, so we cannot yet credit him with the
discovery that SU(2) is two-to-one homomorphic onto SO(3).
7 We set up another correspondence between self-adjoint matrices and 3 vectors in
space (see Sternberg (1994, §1.2) or Goldstein (1980, ch. 4) for details). Let A be a
self-adjoint matrix, and M a member of SU(2). Then conjugation by M, i.e., MAM∗ ,
yields a self-adjoint matrix corresponding to a rotated vector.
8 See also Hadamard (1954, pp. 128–189): ‘Most surprising—I should say
bewildering—facts of that kind are connected with the extraordinary marks of contem-
porary physics. In 1913, Elie Cartan, one of the first among French mathematicians,
thought of a remarkable class of analytic and geometric transformations in relation
to the theory of groups. No reason was seen, at that time, for special consideration
of those transformations except just their esthetic character. Then, some fifteen years
later, experiments revealed to physicists some extraordinary phenomena concerning
electrons, which they could only understand by the help of Cartan’s ideas of 1913.’
9 The limitations on observation made it impossible for Kepler to observe the
perturbations of one planet by another, and he published his ‘laws’ of planetary motion
as though they were not so perturbed. This enabled Newton to derive from these laws
GETTING M O R E O U T O F M AT H E M AT I C S 141
(a) (b)
The fact that the electron ‘lives’ in a two-dimensional complex vector space
and has SU(2) symmetry is detectable even on the macroscopic level. Consider
the carbon-60 molecule, which has the shape of a buckyball, or the truncated
icosahedron, which is made out of hexagons and pentagons like a soccer ball.
The buckyball would appear to have the symmetry I of an icosahedron: sixty
different rotations around its center leave the buckyball invariant.10 Yet if we
want to study the paramagnetic behaviour of the C-60 molecule in a magnetic
field that also has icosahedral symmetry we must ‘pull back’ to the members
of SU(2) that correspond to the members of I: the true symmetry group of the
buckyball, then, is the 120 member subgroup, G, of SU(2) that is two-to-one
homomorphic onto I (Chung et al., 1994, §9).
There is a story within a story here. Laguerre published in 1867 a ‘letter’ to
Hermite in which he defines matrix multiplication and discusses its properties
(Laguerre, 1867: 1898). In the course of this discussion, he introduces matrices
over the integers modulo p, presumably for studies in number theory, but in
any case, not for any physical application. If we consider 2 × 2 matrices, let
p = 5, and restrict ourselves to invertible matrices whose determinant is +1,
we get exactly 120 matrices forming a group, and this group turns out to be
isomorphic to G, a rather startling fact, which means that Laguerre’s number-
theoretic idea turns out to describe the electronic and magnetic properties of
the buckyball—surely we got more out than we put in (Chung et al., 1994, §2)!
Let us now return to the electron itself, which returns to itself only after two
turns. It might be thought that, since the electron ignores standard geometry, it
the inverse square law of gravitational force, and then study the resulting perturbations.
The ‘missing’ lines in the spectrum of hydrogen were actually there (though weak); had
they been detectable (then), a major discovery would have been missed.
10 See Chung and Sternberg (1993) for a nice exposition.
142 M E A N I N G I N M AT H E M AT I C S
could have returned to itself after 3 or any other number of turns. Nevertheless,
there is something ‘natural’ about the number 2 which can be understood using
topology.
Using an example often cited, if you place the end of a belt in a book and
rotate the book a full turn, while holding the other end of the belt, you get a
twist in the belt. But it you turn the book two full turns, you can untwist the
belt by looping it under the book. Topologically this is expressed as follows: if
we make a closed loop of rotations, we cannot ‘shrink’ this loop into a ‘point’
(a point here meaning a constant curve of rotations, a curve that doesn’t rotate
anything). Only a double loop can be so shrunk. This is an indirect fact about
Euclidean space, which is revealed by studying the two-to-one homomorphism
between SU(2) and SO(3) (see Sternberg, 1994, §1.6).
Feynman has another “twist” (excuse the pun) on this idea (see Feynman
and Weinberg, 1987, pp. 56–59). He suggests tying a belt to two electrons,
and get them to switch places. The belt will be found to be twisted, showing
that topologically we have a single rotation, each electron having made what
is equivalent to half a rotation. In that case we should expect the wave function
describing the two electrons to change signs, which is the fundamental prop-
erty of fermions. If this argument is true, and it seems too good to be true, we
get the famous connection between ‘spin’ and ‘statistics’ using SU(2) and a
little topology—no relativity, no field theory. Even if this argument is incon-
clusive, there is a proof connecting SU(2) symmetry with fermionic statistics
using considerations that the inventors of SU(2) could not have dreamed of.
What is most remarkable, perhaps, is the further application of SU(2)
and unitary matrices in general. These further applications disconnect the
concept of unitary matrix more and more from their original application—the
representation of rotations.
I will cite two examples: the first is the application by Heisenberg to
nuclear physics in 1932 of SU(2) symmetry. It turned out that the neutron
and proton are two states of the same particle (called today the ‘nucleon’),
and that, mathematically, the two particles are analogous to the ‘spin up’ and
‘spin down’ states of the electron. The ‘rotation’ that would take the neutron
into the proton and then back to the neutron (a ‘rotation’ that would cause
the wave function of the particle to change sign) cannot be thought of as a
rotation in physical space; nevertheless, the SU(2) symmetry of the nucleon,
called ‘isotopic spin’, has empirical consequences. The physical basis for this
analogy, if there is one, is unknown even today.
When we move to higher-dimensional unitary matrices, the simple con-
nection with rotations breaks down. For example, SU(3) (the group of 3 × 3
unitary matrices with determinant +1), is not homomorphic onto any group
SO(n). Its connection to rotation is only via the analogy to SU(2). Despite
this, or perhaps because of this, SU(3) turns out to describe the three states of
a (type of) quark, which the nucleons are made of (since the nucleons have
integral charge, the quarks have fractional charge, a fact which impeded their
discovery). Rotations turned out to be the tip of a much bigger iceberg.
GETTING M O R E O U T O F M AT H E M AT I C S 143
Another way to generalize the SU(2) two-to-one homomorphism onto the
rotation group was noted by Klein in his lectures on the top (1922, Lecture I).
If we drop the requirement of unitariness, and consider only 2 × 2 complex
matrices with determinant 1, we get the group of matrices today known as
SL(2,C). Klein shows (see p. 626 for a statement of this theorem) that there is
a two-to-one homomorphism between SL(2,C) and a group of transformations
on a two-dimensional manifold, and this group turns out to be isomorphic to
the proper Lorentz group L◦ (the group of Lorentz transformations that have
positive determinants and preserve the forward ‘light cone’), a homomorphism
which has itself great physical significance when we are studying relativistic
quantum theory.
I think that the applicability of ‘magical’ (Penrose’s phrase) complex num-
bers to rotations and its generalization to SU(2) and beyond is not unusual
in the history of mathematics, but it well illustrates the phenomenon that
mathematical concepts tend to have latent information, which can be used in
the development of mathematics. What is more, nature seems to make use of
these mathematical possibilities.
There are three elements coming together here: mathematics, nature, and
the human mind. Which of these three is responsible for the remarkable
richness of ‘mathematical essences’?
Acknowledgements
I would like to thank Shlomo Sternberg for crucial aid and advice. Thanks also
to Shmuel Elitzur, Sylvain Cappell, and Carl Posy for valuable discussions
over the course of many years. The research that went into this article was
supported by the Israel Science Foundation, Grant no. 251/06, and I am very
grateful for this support.
Comment on Mark Steiner’s ‘Getting
more out of mathematics than what
we put in’
Marcus du Sautoy
The idea of getting more out than you put in is one of the most attractive
features of my subject. I think this quotient is probably higher for mathematics
than for any of the other sciences, which is one reason I chose mathematics
over the mess of biology. Maybe it’s also because I have such a bad memory.
You only need a few axioms and everything else starts spilling out once you
start. Biology and chemistry always seemed to require you to memorize the
things Nature selected, which at times appear quite arbitrary.
The art of the mathematician is often to pick the rules of the game to
maximize this quotient. Although Steiner makes a distinction between games
and maths, I think there is the same satisfaction in seeing how the simple rules
of ‘Go’ lead to a game with such a rich array of plays, and the way three simple
axioms for the definition of a group can lead to the Monster simple group and
all the Lie groups that underpin physics.
Which brings me to the challenging point raised by Steiner’s article: the
extraordinary synergy between the abstract, beautiful world of mathematics
and the physical, messy world of physics, chemistry and biology. It is perhaps
bizarre that a perfect circle or right-angled triangle, the most fundamental
objects of mathematics, may not have any physical reality if quantum physics is
right about the world being quantized into discrete bits. And the infinities that
I bandy around with ease might have no concrete realization in a universe that
is potentially finite in nature. Nevertheless it is astounding how these objects
of the mind help us to predict the future behaviour of our messy universe. How
come the imaginary numbers we create to solve a cubic equation are the same
numbers that are crucial to describing the quantum world?
Perhaps it is an anthropomorphic answer as Steiner suggests in response
to my article. We make choices about the maths we like to celebrate. Where
did that maths and the excitement come from originally? From describing the
GETTING M O R E O U T O F M AT H E M AT I C S : C O M M E N T 145
physical world. The Egyptians wanted to know the volume of a pyramid. They
needed to know after all how many bricks to use. But to calculate the volume
they are led to the discovery of the power of cutting a shape into infinitely
many, infinitely thin pieces, which they can rearrange to make the problem
easier. An early form of integral calculus at work.
The process of cutting a real pyramid like this is clearly absurd on a practi-
cal level, yet a projection has been established from the world of mathematics
down onto our messy world. But because the world of mathematics began
it’s journey trying to describe and predict physical reality, perhaps it isn’t
so unexpected that the maths we generate in a purely abstract form, and for
its intrinsically internal fascination, nevertheless can often find itself being
projected back down to our messy universe generations after the journey was
kicked off.
A last point. Sometimes maths is very good at showing why you can’t get
any more out from what you put in. Real numbers led to complex numbers led
to quaternions and gave birth to octonions, but then mathematicians can prove
that you’re not going to get any more out of this. Similarly, the Lie groups E6 ,
E7 and E8 are such beautifully powerful structures, but the maths shows why it
stops there. There can’t be an E9 . Sometimes you get less out than you might
expect. But knowing that is sometimes as exciting as getting lots out of a small
investment. The exceptional Lie groups are special because of their unique
character. Still, it is amazing that E8 could be the model for the fundamental
particles that make up the fabric of reality. Nature certainly has good taste.
This page intentionally left blank
References